From Genotype to Phased Haplotype: Multiplex Digital PCR-Based Haplotyping at the APOE Locus in Alzheimer’s Disease

From Genotype to Phased Haplotype: Multiplex Digital PCR-Based Haplotyping at the APOE Locus in Alzheimer’s Disease | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article From Genotype to Phased Haplotype: Multiplex Digital PCR-Based Haplotyping at the APOE Locus in Alzheimer’s Disease Sunny Chen, Eun-Gyung Lee, Lesley Leong, Jessica Tulloch, Chang-En Yu This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7634101/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Backgrounds: Alzheimer’s disease (AD) risk reflects multi-locus variant effects. Despite advances in genotyping techniques, current molecular haplotyping approaches involve intricate operational frameworks and incur high implementation costs, and most studies therefore rely on unphased genotypes. We developed a straightforward approach to reconstruct locus-wide haplotype structure within the APOE locus and assess for potential AD-risk haplotypes. Methods We analyzed 49 postmortem brain samples (40 AD and 9 Controls) heterozygous for APOE ε3/ε4. We developed a Multiplex Digital PCR assay to resolve allelic phase configurations of AD-associated SNPs across three closely linked loci within the APOE genomic region: TOMM40 (rs2075650), APOE (rs429358) and APOC1 (rs12721046). Gaussian mixture modeling was used to reconstruct sample-level phased haplotype structure. Results We achieved high-confidence reconstruction of three-SNP haplotypes spanning a 25.6 kilobase (kb) region. The diplotype frequencies did not differ significantly in AD and Control groups (G = 11.36, p = 0.25). A suggestive trend was observed between the AD and Control groups for TOMM40 rs2075650-A and APOC1 rs12721046-G haplotype (G = 6.50, p = 0.09). Conclusions Although underpowered and not statistically significant, this proof-of-concept shows that multiplex digital PCR can support simple, sample-level haplotype phasing at the APOE locus. This strategy provides a robust platform for mechanistic investigation and translational application by integrating phased haplotype configurations with three-dimensional chromatin architecture–associated regulatory dynamics, thereby informing locus-specific therapeutic targeting. APOE dPCR Genotype Phased Haplotype Alzheimer’s Disease Figures Figure 1 Figure 2 Introduction Complex disorders such as late-onset Alzheimer's disease (LOAD) are characterized by substantial genetic contributions. Large-scale genome-wide association studies (GWAS) and next-generation sequencing approaches ( 1 , 2 ) have identified over 150 loci associated with LOAD, the majority of which confer modest risk, with reported odds ratios ranging from approximately 1.1 to 1.5. Among these, the apolipoprotein E gene ( APOE ) locus exhibits the most pronounced effect on disease susceptibility, with an estimated odds ratio of 3.7 ( 3 ), markedly surpassing that of other loci. This elevated genetic risk underscores the pivotal role of APOE in modulating LOAD pathogenesis. Elucidating the genetic architecture of the APOE locus may inform more precise therapeutic targets for LOAD. Multiple independent GWAS have reported robust associations between single nucleotide polymorphisms (SNPs) across three adjacent genes— TOMM40 , APOE , and APOC1 —and LOAD ( 1 , 4 ) in the APOE locus. This shared signal may reflect one of two potential underlying mechanisms: [1] the association is primarily driven by APOE , with signals from neighboring loci attributable to strong linkage disequilibrium (LD) ( 5 ); [2] a combinatorial effect wherein variants across all three genes jointly influence disease susceptibility and related phenotypes. While distinguishing between these hypotheses remains methodologically challenging, the latter scenario introduces a compelling framework in which coordinated regulatory or functional elements across the locus exert synergistic effects on AD pathogenesis. This alternative scenario is supported by biological evidence, as characteristic AD phenotypes frequently encompass mitochondrial dysfunction and impaired innate immune responses—features that may be mechanistically linked to the adjacent genes flanking APOE . Mitochondrial dysfunction is one of the hallmarks of AD ( 6 ); reduced Tom40—encoded by TOMM40 , a core component of the mitochondrial translocase complex—compromises mitochondrial integrity ( 7 ), promotes metabolic dysregulation, oxidative stress, and mitochondrial damages ( 8 ). APOC1 encodes apolipoprotein C1 involved in lipid metabolism ( 9 ) and innate immune responses ( 10 , 11 ); increased brain apoC1 level and risk-associated APOC1 genetic variants have also been linked to increased AD risk ( 12 ). Elevated expression of TOMM40 , APOE , and APOC1 is observed in human cellular models under oxidative stress, and in postmortem brain tissues from AD patients ( 13 ). To evaluate this model, it is critical to define haplotype structures associated with differential risk and to develop streamlined molecular haplotyping methodologies capable of resolving locus-level genetic architecture with high fidelity. The genetic study of SNPs has been an important tool for scientists to identify common genetic variants associated with the risk of diseases ( 14 ). Most disease-associated SNPs identified by GWAS are not located within genes or well-defined regulatory elements, making the functional assignment of these SNPs challenging. Through computational algorithms, studies have uncovered non-random associations of SNP alleles at different sites that are in LD ( 15 ). These alleles that are in LD are frequently co-inherited and can vary between different ethnic populations; the combinations of the alleles located on the same chromosome are known as haplotypes ( 16 ). Despite recent advancements in genotyping technologies that have made SNP pattern characterization more efficient, most studies have not accounted for the phased haplotype (i.e., physically separated maternal and paternal chromosomes) of the DNA samples ( 17 ). Phased haplotype analysis is essential for studying complex diseases. When multiple disease-associated genetic alleles are in LD with each other, it becomes difficult to determine if these alleles collectively influence disease risk. This complexity is further compounded by whether the alleles are located on the same or different chromosomes (i.e., paternal and maternal). In a hypothetical scenario where a person inherited heterozygote major and minor allele in both SNP1 (M 1 /m 1 ) and SNP2 (M 2 /m 2 ) can yield 4 possible haplotype patterns: M 1 M 2, M 1 m 2, ) m 1 M 2 and m 1 m 2 . The possible combinations of haplotype patterns grow exponentially as more SNP sites are added into the study. As a result, deciphering the molecular mechanisms underlying disease-associated SNPs remains challenging. This leads to phase uncertainty, which diminishes the power to accurately detect specific genetic loci or combinations of alleles at different loci that may contribute to disease risk signals. On the contrary, information gained from phased haplotype can clearly distinguish the independent or additive effects of linked allele variants. This helps to determine whether the risk of disease arises from a single locus or from a combination of multiple loci, which highlights the importance of phased haplotype analysis in complex diseases. The main challenges of phased haplotyping stem from its computational complexity and the absence of cost-effective methods ( 17 ). Commonly used SNP analysis techniques such as TaqMan-based genotyping and SNP microarrays do not provide sufficient information to infer the haplotype pattern when both SNP sites are heterozygous for both alleles. Although frequently used sequencing methods such as Sanger sequencing and next-generation sequencing can deliver clear sequence reads of genotype data, the information from these reads is insufficient to phase haplotypes. Although there are higher-fidelity approaches that do not rely on algorithmic inference—such as nanopore sequencing, which is capable of effectively identifying haplotype structures within 25 kilobases ( 18 ), and clone-based systematic haplotyping, which can reconstruct haplotype of regions spanning more than 60 kilobases ( 19 )—, these techniques require intensive trainings, involve tedious works, are extremely expensive, and cannot process large sample sizes simultaneously. Consequently, available haplotype phasing protocols are often expensive, rely on large-scale sequencing for haplotype reconstruction, or require long and stringent process of computational assembly ( 20 , 21 ). The most commonly used approach applies statistical algorithms, which only estimate haplotypes with chromosome phase uncertainty and do not accurately reconstruct the original haplotypes. For example, known algorithmic methods such as IMPUTE2 ( 22 ), MaCH ( 23 ), PHASE ( 24 ), fastPHASE ( 25 ), and Shape-IT ( 26 ) require extremely large sample sizes for accuracy, take a long time to run, and use varying computational standards for haplotype reconstruction. Furthermore, there is no gold standard on how to perform statistical-based haplotype inference. These limitations make current techniques extremely difficult to apply into clinical setting. In this study, we have developed a new molecular haplotyping procedure to define the phased haplotypes using the digital PCR (dPCR). The goal is to offer a streamlined approach that simplifies the processing time required for haplotype reconstruction, and facilitate haplotype studies for complex diseases. Unlike existing methodologies, this haplotyping protocol is not restricted by the distance between SNP sites, is less expensive compared to sequencing, is capable of processing multiple samples simultaneously, and does not require any statistical computation to infer haplotype structure. This makes it an attractive method for conducting phased haplotype studies. To develop this haplotyping procedure, we concentrated on the haplotype structure of the human APOE gene locus, as multiple SNPs across three genes ( TOMM40 , APOE , and APOC1 ) in this region have been consistently linked to LOAD by GWAS ( 1 , 27 ). Among these SNPs, TOMM40 rs2075650, APOE rs429358; and APOC1 rs12721046 exhibit the strongest association signals on AD risk ( 27 – 30 ), although their effects vary between ancestries due to inherited haplotype structure (31). For example, the ε4 variant (characterized by non-synonymous SNPs rs429358 and rs7412) of the gene APOE is associated with increased risk of development of AD ( 32 , 33 ). However, multiple studies have suggested that despite the high prevalence rate of APOE ε4 variant in African and Hispanic populations, the impact of the risk alleles on AD risk is several folds lower than Caucasian population ( 34 – 36 ). The reason for such discrepancy may be due to different ancestral haplotype structures and combination of genetic alleles, which lead to diverse effects of AD risk. A potential explanation is that the risk effect of APOE ε4 allele can be further modified by adjacent genetic variants across different populational groups. Therefore, these SNPs serve as excellent examples for phased haplotype construction. By reconstructing phased haplotypes from discrete allelic variants within the APOE locus, we aim to elucidate the locus-specific genetic architecture contributing to AD susceptibility Methodology DNA Extraction : Genomic DNA (gDNA) were isolated from frozen post-mortem brain (PMB) tissues from the cerebellum and frontal cortex. The PMB was lysed with TissueLyser L1, 40Hz, 1 min (Qiagen). The gDNA were extracted using the AllPrep DNA/RNA Mini Kit (Qiagen). All DNA isolation procedures were performed according to the manufacturer’s protocols. Nucleic acid concentrations and qualities were assessed by NanoPhotometer (Implen), and the samples were stored at -20 °C prior to use. Evidence-Based Confirmation of SNPs Candidate To identify meaningful SNPs that may influence regulatory elements within a TAD region, we used bioinformatics tools to prioritize functional SNP variants for future phased haplotype studies. One such tool is the FORGEdb scoring system (37), which integrates multiple functional evidences. We suggest selecting SNPs with a high FORGEdb score (≥8), as this indicates strong support from functional evidences such as activity-by-contact (ABC) interaction, chromatin looping, DNase I hotspot, expression quantitative trait locus (eQTL) association, histone mark, or overlap with transcription factor (TF) motifs. As an additional criterion, we recommend incorporating the RegulomeDB scoring system (38) as a secondary validation step. SNPs with a low RegulomeDB score (1–5) will ensure that the result was consistent between both databases, and there was at least one evidence of DNase peak or TF binding at the SNP locations. For example, the SNPs selected in this study—rs2075650 (FORGEdb score = 8, RegulomeDB score = 1b), rs429358 (FORGEdb score = 8, RegulomeDB score = 4), and rs12721046 (FORGEdb score = 8, RegulomeDB score = 5)—meet these criteria. Haplotyping Control Templates (HCTs) : Eight artificially synthesized 467 base pairs double stranded chimeric DNA fragment containing genetic sequences surrounding the SNPs TOMM40 rs2075650, APOE rs429358, and APOC1 rs12721046 (Integrated DNA Technologies). The NCBI36/hg18 genomic coordinates of the conjoined sequences included: TOMM40 Chr19:50087357–50087503; APOE Chr19:50103719–50103851; and APOC1 Chr19:50113012–50113198 (Supplement Figure S1). Sample Selection and Genotyping of TOMM40, APOE and APOC1 : The gDNA isolated from PMB was used for genotyping. The TOMM40 SNP, rs2075650 were genotyped using the TaqMan allele discrimination assay C___3084828_20 (ThermoFisher). The APOE SNP, rs429358 were genotyped using the TaqMan allele discrimination assay C___3084793_20 (ThermoFisher). The APOC1 SNP, rs12721046 was genotyped using the TaqMan allele discrimination assay C__31478296_10 (ThermoFisher). dPCR Sample Preparation : A total of 10 µl of gDNA isolated from post-mortem brain tissues (1µg concentration) was mixed with 90 µl of Tris-EDTA Buffer Solution, pH 7.4 (Sigma-Aldrich) to a final concentration of 10 ng/µl and a total volume of 100 µl. The diluted DNA sample was placed in a 0.2 µl Bioruptor microtubes (Diagenode) and sheared with Bioruptor UCD-200 Sonication System (Diagenode) in 4°C water bath for 4 cycles of 1 second on/30 seconds off sonication cycles at high settings. In the absence of validated calculation tool, we estimated the mean fragment length by scaling from Diagenode’s Bioruptor Standard benchmark, with the assumption that the fragment length is inversely proportional to total ON time. Based on the specific parameters, 8 cycles at 30 seconds on/30 seconds off (total; ON = 240 seconds) yield ~350 bp product, whereas 30 cycles (total ON = 900 seconds) yield ~200 bp products (39). We estimated the length to be ≈ , which is approximately 21–45 kilobases. The sample was transferred into a 1.5 ml microcentrifuge tube (Eppendorf), and was stored in -20 °C. A detailed rationale for preferring sonication over restriction enzyme digestion is provided in Supplementary Methods Section 1 . dPCR Primer and Probe Design : The Absolute Q Digital PCR System (ThermoFisher) used TaqMan-based dPCR method to determine the haplotype information of each sample. The assays were designed with the PrimerQuest Tool (Integrated DNA Technology) to ensure for the best quality. For best dPCR amplification efficiency, the expected size of the amplicon was set to be close to 100 base pairs, and the maximum size should not exceed 150 base pairs. Since Absolute Q Digital PCR System (ThermoFisher) used ROX dye as passive reference, any fluorescent dyes with excitation spectrum of 580 ± 10 nm and emission spectrum of 623 ± 14 nm were not be selected for probe design. Information on dPCR primers and probes is listed in Supplement Table S1. The challenges encountered and the methods used for assay optimization are described in Supplementary Methods Section 2 . dPCR Haplotyping Protocol : Each dPCR reaction included 3 µl of sonicated gDNA input (30 ng), 2 µl of 5X Absolute Q DNA Digital PCR Master Mix (ThermoFisher), 1 µl of 10X digital PCR assay for each candidate SNPs, and add nuclease-free water up to total volume of 10 µl. A total of 9 µl of the dPCR mixture was loaded into the QuantStudio Absolute Q MAP 16 Digital PCR Plate (Applied Biosystems, ThermoFisher), and 15 µl of Absolute Q Isolation Buffer (Applied Biosystems, ThermoFisher) was added on top of the reaction to avoid cross-contamination and evaporation. The concentration of APOE rs429358 assay was increased to 20X due to lower primer/probe efficiency. The thermal cycling profile for dPCR genotyping of TOMM40 rs2075650 and APOE rs429358 was 10 min at 96ºC, followed by 40 cycles of 5 sec at 96ºC and 15 sec at 62ºC. The thermal cycling profile for dPCR genotyping of APOE rs429358 and APOC1 rs12721046 was 10 min at 96ºC, followed by 40 cycles of 5 sec at 96ºC and 15 sec at 66ºC.The genotype result was analyzed with QuantStudio Absolute Q Digital PCR Software 6 (Applied Biosystems, ThermoFisher), where the fluorescence signals of the targeting SNPs were translated into partitions on a 2D-plot. Phase Haplotype and Statistical Analyses : The data were analyzed using Python Version 3.13.5 (40) for Windows. Data processing and Gaussian Mixture Models (GMM) analyses were performed using built-in and open-source libraries, including pandas (41), NumPy (42), and scikit-learn (43). The haplotype data were analyzed using R-Program Version R-4.5.1. The statistical package DescTools (CRAN—Package DescTools (r-project.org)) was used to perform G-Test as a likelihood ratio test to differentiate between high-risk and low-risk haplotype patterns. Standardized residuals from Chi-Square were used to determine the haplotype that has the largest deviation. Post-hoc 2x2 Fisher tests were used to determine whether or not a specific haplotype statistical significantly contribute to the difference between AD and CTRL. Results 1. Development of a Multiplexing dPCR (mdPCR) Procedure for Molecular Haplotyping 1.1. Selection of SNPs and Design of SNP Assays We initially selected three candidate tagging SNPs within the APOE locus—each previously demonstrated to be robustly associated with LOAD risk and prioritized using bioinformatic functional annotations (FORGEdb, RegulomeDB; see Methods, Evidence-based confirmation of candidate SNPs )—to develop mdPCR procedure. These SNPs span three genes ( TOMM40 [rs2075650], APOE [rs429358], and APOC1 [rs12721046]) and cover a 25.6 kb genomic region (see Fig. 1 a). We then designed TaqMan-based SNP assays (rs2075650 g-allele, rs429358 c-allele, and rs12721046 G-allele) using PrimerQuest Tool (Integrated DNA Technology). Each probe was tagged with various reporter dyes. To streamline our research and development process, we also designed and synthesized artificial chimeric DNA templates, referred to as Haplotype Control Templates (HCT). These HCT artificially combine three tagging SNPs and their flanking sequences into a condensed 467 bp DNA template. Each HCT was designed to contain a unique allele of the three SNPs (see Figure S1 ). Consequently, a total of eight HCTs were synthesized. Each HCT was diluted and subjected to copy number quantification, permitting the deposition of individual HCT molecule into distinct dPCR partitions. It was then used in subsequent multiplexing dPCR reactions to test compatibility of assays and signals as well as validate the instruments’ accuracies in detecting copy numbers. 1.2. Testing the Compatibilities of SNP Assays, Signals, and dPCR Instruments using synthetic HCT DNA constructs To determine the most suitable instrument for the mdPCR procedure, we conducted an initial test to compare the signal sensitivity and multiplex capability of the two digital PCR systems: the QIAcuity dPCR System (Qiagen) (hereinafter referred to as “QIAcuity”) and the QuantStudio Absolute Q dPCR System (ThermoFisher) (hereinafter referred to as “Absolute Q”). To make use of the full spectrum of commercially available fluorescent dyes, we evaluated signals from various fluorescent probe combinations spanning the lowest to highest emission maxima—FAM, VIC, HEX, ABY, TAMRA, ROX, Texas Red, and Cy5—covering approximately 520–670 nanometers (nm) (emission maxima; FAM ~ 520 nm to Cy5 ~ 670 nm). We observed that the Absolute Q instrument consistently produced better and reproducible results. While we do not observe spectral overlaps between the dyes, the fluorescent signal from Cy5 (crimson channel; emission maxima ~ 667–670 nm) was consistently lower than dyes from the lower excitation/emission spectra. Therefore, when multiplexing SNP assays, the assay with lowest PCR efficiency should avoid far-red labeling, particularly with Cy5 or adjacent wavelengths. The most effective duplex combinations of fluorescent probes were FAM with VIC or HEX, FAM with ABY or TAMRA, and VIC or HEX with ABY or TAMRA. Detailed evaluations of dPCR platform capabilities are provided in Supplementary Methods Section 3 . 1.3. Interpretation of Signals and Partitions from Output of dPCR Instrument Upon completion of dPCR reaction, the Absolute Q software automatically converted the fluorescent signals from partition (or microwell) counts from each reaction into 2D plots. The partitions were separated into 4 distinct quadrants based on the fluorescent signals detected from each partition in the dPCR plate. A representative duplex reaction using FAM and VIC probes is shown in Supplement Figure S2. The upper left corner of the 2D plot is the (+,–) quadrant, which tallied the partitions that emitted only the FAM fluorescent signal. The bottom right corner is the (–,+) quadrant that counted the partitions that emitted only VIC fluorescent signal. The upper right corner is the (+,+) quadrant that totaled the partitions that emitted both FAM and VIC signals. The final (–,–) quadrant located at the bottom left corner of the plot displayed partitions that did not emit any fluorescent signals. At optimal dilution, the concentration of DNA is sufficiently low such that each partition within the dPCR assay is statistically occupied by no more than a single template molecule. In theory, if the specific alleles of both target SNPs are located on the same DNA template, both fluorescent signals will be detected from the same partition (see Figure S2). The (+,+) quadrant will show elevated positive fluorescent signals (see Figure S2a). Conversely, if the specific alleles of the target SNPs are located on separate DNA templates, only a single fluorescent signal will be emitted from one partition at a time. This will lead to an increase in positive partitions in the (+,–) and (–,+) quadrants (see Figure S2b). Thus, this mdPCR method enables the accurate determination of haplotype structures. Following successful optimization of the protocol using synthetic HCT DNA constructs, the procedure was subsequently implemented for genomic DNA analysis. 2. Validation of mdPCR in genomic DNA Haplotyping 2.1. Selection and Preparation of Genomic DNA We first genotyped and selected a panel of PMB samples that were heterozygous at the loci of interest but could not be phased using standard TaqMan genotyping procedure. Because all sites were heterozygous, the genotype data lacked phase information, precluding unambiguous inference of haplotypes (allele co-segregation across loci). Table 1 provides the demographic information of the selected PMB samples in this study. Given the substantial molecular mass and volumetric complexity of human genomic DNA, fragmentation into smaller units is necessary to facilitate stochastic partitioning within the digital PCR array. In the standard dPCR protocol, the manufacturer recommended users to reduce the size of the input gDNA by digesting them with 1 Unit of 6-base cutting restriction enzymes. Based on this recommendation, we tested various restriction enzymes that do not cleave our targeted region to digest gDNAs prior to mdPCR. However, this approach consistently resulted in a low number of positively signaled partitions, making it difficult to differentiate between quadrant-related partition counts. To address this technical limitation, we implemented an alternative approach utilizing sonication to achieve controlled fragmentation of gDNA. Table 1 Sample Demographic Subjects AD CTRL Sample number_n 40 9 Sex-Female_n (%) 23 (57.50%) 4 (44.44%) Age at death_mean (SD) 84.05 (9.15) 85.78 (5.93) Age at onset_mean (SD) 74.95 (9.38) – Disease Duration_mean (SD) 10.15 (4.73) – Postmortem interval_mean hour (SD) 4.99 (1.67) 5.72 (3.34) CERAD Score_n Absent – 1 Sparse – 3 Moderate 3 5 Frequent 37 – BRAAK Stage_n I – – II – 1 III – 8 IV – – V 9 – VI 31 – Using Diagenode sonicator to optimize the sonication protocol, we prepared multiple aliquots of the same gDNA sample, and subjected them to sonication under varying energy conditions. Initially, we performed 4 cycles of sonication protocol with 5 different energy levels (1 to 5 seconds on/30 seconds off). We compared the number of positive partitions of the signals produced by TaqMan FAM assay for APOE SNP rs429358 across samples treated at each sonication energy level. The results indicated that sonication energy at 1 second on/30 seconds off produced the best outcome. After defining the optimal energy, we then tested for the optimal number of cycles achieve best dPCR result. Using the same setting, we compared the dPCR outcomes from samples subjected to 2, 4, 6, and 8 sonication cycles. The results showed that the highest partition count was achieved with the 4 cycles protocol. Based on this parameter, the average gDNA fragment size was estimated to range between 21 and 45 kilobases. The detailed calculation methodology is provided in the “dPCR Sample Preparation” subsection of the Materials and Methods. 2.2. Haplotyping SNPs rs2075650 (TOMM40) rs429358 (APOE) and rs12721046 (APOC1) with mdPCR A panel of PMB samples carrying the APOE ε3/ε4 genotype was selected and genotyped using commercial TaqMan allelic discrimination assays to determine allelic composition at SNPs rs2075650 and rs12721046. Samples exhibiting homozygosity at both SNP loci were excluded from subsequent molecular haplotyping, as their haplotypes could be confidently inferred without the need for further analysis. Among the remaining samples, we identified a subset that were heterozygous at rs2075650, rs12721046, or both loci. For these heterozygous samples—where phase could not be inferred in silico —we employed mdPCR to determine the cis- or trans-configurations of the alleles coexisting at the single-haplotype level. Based on genomic coordinates, we selected APOE SNP rs429358 as the anchoring SNP. For each selected PMB sample heterozygous at rs2075650, rs12721046, or both loci, we conducted either one or two mdPCR reactions: one reaction for rs2075650-g-allele and rs429358-c-allele, and another for rs429358-c-allele and rs12721046-G-allele (Fig. 1 b). Optimized mdPCR output profiles are depicted in Fig. 2 . Figure 2 a shows the 2D plot of a DNA sample that is homozygous at both target sites (rs429358 c/c and rs12721046 G/G), representing two copies of the c–G haplotype. This sample was used as an internal control for the mdPCR protocol to improve the clustering accuracy of the subsequent Gaussian Mixture Models (GMM) analysis (see next section). Figure 2 b depicts the 2D plot of a sample that is heterozygous on both SNPs, rs429358 (T/c) and rs12721046 (G/a), carrying a single copy of the c–G haplotype. The plot shows positive but reduced signals in the (+,+) quadrant, indicating the presence of a single c-G Cis haplotype. In contrast, the sample in Fig. 2 c exhibits a distinct 2D plot pattern of a DNA sample that is heterozygous on both SNPs, with minimal signal in the (+,+) quadrant. This suggests the absence of a c-G haplotype, supporting a Trans haplotype structure—specifically, c-a and T-G. We utilized the mdPCR technique and performed three independent runs for each PMB sample, generating three experimental replicates per sample. The raw partition numbers from each quadrant of the respective 2D plots were extracted and are reported in Supplement Table S2. While partition counts within the (+,+) quadrant offer a preliminary estimate of c-G haplotype copy number (e.g., 0, 1, or 2), data from three independent experimental runs revealed substantial variability, limiting the reliability of this direct approach of copy number calling. 2.3. Interpreting mdPCR Data To improve confidence in copy number inference, we employed more robust statistical algorithms, Gaussian Mixture Models (GMM) ( 44 , 45 ), which perform pattern recognition analysis and capable of accommodating assay noise and partitioning uncertainty. This approach streamlines the interpretation of mdPCR outputs and eliminates the need for manual inspection of 2D plot signals. The GMM assumes that the data are generated from a mixture of Gaussian distributions (i.e., multiple clusters), and uses the Expectation-Maximization (EM) algorithm ( 46 ) to estimate the posterior probabilities that each data point belongs to a specific cluster. The results from the three experimental replicates were combined to account for plate-to-plate variation. We extracted all partition numbers from the (+,+), (+,–) and (–,+) quadrants to calculate for the fractions for the Cis and Trans haplotypes of each sample. The Cis Fraction was calculated as \(\:\varvec{C}\varvec{i}\varvec{s}\:\varvec{F}\varvec{r}\varvec{a}\varvec{c}\varvec{t}\varvec{i}\varvec{o}\varvec{n}=\:\frac{{N}_{(+,+)}}{{N}_{(+,-)}+{N}_{(-,+)}+{N}_{(+,+)}\:}\) , and the Trans Fraction as \(\:\varvec{T}\varvec{r}\varvec{a}\varvec{n}\varvec{s}\:\varvec{F}\varvec{r}\varvec{a}\varvec{c}\varvec{t}\varvec{i}\varvec{o}\varvec{n}=\:\frac{{N}_{(+,-)}+{N}_{(-,+)}}{{N}_{(+,-)}+{N}_{(-,+)}+{N}_{(+,+)}\:}\) . These fractions were then used to estimate the posterior probabilities for assigning each sample to the appropriate Cis or Trans category. Since input DNA and assay concentrations were controlled, we exclude the (–,–) quadrant—which contains negative or non-informative partitions—from the analysis. The GMM analyses for TOMM40 – APOE haplotype (rs2075650–rs429358) and APOE – APOC1 haplotype (rs429358–rs12721046) were presented in Supplement Table S3. This approach enabled more efficient and accurate haplotype resolution, enhancing confidence in variant assignment. 2.4. Reconstruction and Identification of Risk Haplotype Structure For samples classified as Cis-configuration for the TOMM40 – APOE haplotype (rs2075650–rs429358), the deduced haplotype is g–c; whereas samples classified as Trans-configuration have the A–c haplotype. For samples evaluated as Cis for the APOE – APOC1 haplotype (rs429358–rs12721046), the inferred haplotype is c–G; whereas Trans samples have the c–a haplotype. Integrating both sets of pairwise haplotype data from Supplement Table S3 enabled reconstruction of the TOMM40 - APOE - APOC1 haplotype architecture for each sample (Fig. 1 d). The corresponding diplotypes and haplotypic configurations are detailed in Table 2 . Table 2 Haplotype Patterns in APOE ε3/ε4 PMB Samples SampleID rs2075650 rs425358 rs12721046 T40- APOE APOE- APOC1 Diplotype ε3 Haplotype ε4 Haplotype T40-APOC1 Haplotype 1 T40-APOC1 Haplotype 2 AD1 Ag Tc Ga TRANS TRANS gTG|Aca gTG Aca gG Aa AD2 AA Tc Ga - TRANS ATG|Aca ATG Aca AG Aa AD3 AA Tc Ga - CIS Ata|AcG Ata AcG Aa AG AD4 AA Tc GG - - ATG|AcG ATG AcG AG AG AD5 Ag Tc Ga TRANS TRANS gTG|Aca gTG Aca gG Aa AD6 AA Tc GG - - ATG|AcG ATG AcG AG AG AD7 AA Tc GG - - ATG|AcG ATG AcG AG AG AD8 Ag Tc Ga TRANS TRANS gTG|Aca gTG Aca gG Aa AD9 AA Tc GG - - ATG|AcG ATG AcG AG AG AD10 Ag Tc Ga TRANS CIS gTa|AcG gTa AcG ga AG AD11 Ag Tc Ga CIS CIS ATa|gcG ATa gcG Aa gG AD12 Ag Tc Ga TRANS TRANS gTG|Aca gTG Aca gG Aa AD13 Ag Tc Ga CIS CIS ATa|gcG ATa gcG Aa gG AD14 Ag Tc Ga CIS CIS ATa|gcG ATa gcG Aa gG AD15 Ag Tc Ga CIS CIS ATa|gcG ATa gcG Aa gG AD16 Ag Tc Ga CIS CIS ATa|gcG ATa gcG Aa gG AD17 AA Tc GG - - ATG|AcG ATG AcG AG AG AD18 Ag Tc Ga CIS CIS ATa|gcG ATa gcG Aa gG AD19 Ag Tc Ga TRANS CIS gTa|AcG gTa AcG ga AG AD20 Ag Tc Ga TRANS CIS gTa|AcG gTa AcG ga AG AD21 Ag Tc Ga TRANS TRANS gTG|Aca gTG Aca gG Aa AD22 Ag Tc Ga CIS CIS ATa|gcG ATa gcG Aa gG AD23 Ag Tc Ga TRANS TRANS gTG|Aca gTG Aca gG Aa AD24 AA Tc Ga - CIS ATa|AcG ATa AcG Aa AG AD25 Ag Tc Ga CIS CIS ATa|gcG ATa gcG Aa gG AD26 AA Tc GG - - ATG|AcG ATG AcG AG AG AD27 Ag Tc Ga CIS CIS ATa|gcG ATa gcG Aa gG AD28 Ag Tc Ga TRANS TRANS gTG|Aca gTG Aca gG Aa AD29 Ag Tc Ga TRANS TRANS gTG|Aca gTG Aca gG Aa AD30 AA Tc GG - - ATG|AcG ATG AcG AG AG AD31 Ag Tc Ga CIS TRANS ATG|gca ATG gca AG ga AD32 Ag Tc GG TRANS - gTG|AcG gTG AcG gG AG AD33 AA Tc GG - - ATG|AcG ATG AcG AG AG AD34 Ag Tc GG CIS - ATG|gcG ATG gcG AG gG AD35 Ag Tc GG TRANS - gTG|AcG gTG AcG gG AG AD36 Ag Tc GG TRANS - gTG|AcG gTG AcG gG AG AD37 Ag Tc GG CIS - ATG|gcG ATG gcG AG gG AD38 AA Tc GG - - ATG|AcG ATG AcG AG AG AD39 AA Tc GG - - ATG|AcG ATG AcG AG AG AD40 Ag Tc Ga TRANS TRANS gTG|Aca gTG Aca gG Aa CTRL1 Ag Tc Ga TRANS TRANS gTG|Aca gTG Aca gG Aa CTRL2 Ag Tc Ga CIS CIS ATa|gcG ATa gcG Aa gG CTRL3 Ag Tc Ga TRANS TRANS gTG|Aca gTG Aca gG Aa CTRL4 AA Tc GG - - ATG|AcG ATG AcG AG AG CTRL5 Ag Tc Ga CIS CIS ATa|gcG ATa gcG Aa gG CTRL6 Ag Tc aa CIS - ATa|gca ATa gca Aa ga CTRL7 Ag Tc Ga TRANS TRANS gTG|Aca gTG Aca gG Aa CTRL8 Ag Tc Ga TRANS TRANS gTG|Aca gTG Aca gG Aa CTRL9 Ag Tc Ga TRANS TRANS gTG|Aca gTG Aca gG Aa *** INSERT Table 2 HERE *** Following reconstruction of the precise molecular haplotype structures of each sample, we subsequently proceeded to evaluate the contribution of these haplotypes and diplotype configurations to AD risk. The initial G-Test revealed no statistically significant differences between the diplotypes in AD and Control groups (G = 11.36, p-value = 0.25). Stratifying the samples by APOE ε3 and ε4 status suggested a possible trend among ε4 carriers, although these results did not reach statistical significance (ε3: G = 4.24, p-value = 0.24; ε4: G = 5.87, p-value = 0.12). To mitigate potential type II error due to limited sample size, we focused our downstream analysis on the most informative variants— TOMM40 rs2075650 and APOC1 rs12721046—and combined the data accordingly. This approach revealed a borderline significant difference between the AD and Control groups for the A-G haplotype (G = 6.50, p-value = 0.09). To further explore this signal, we conducted a Chi-square analysis of standardized residuals, which confirmed that the A-G haplotype showed the largest deviation. However, the result remained below the threshold for statistical significance (|Standardized Residual of A-G| = 1.70). A post hoc Fisher’s exact test yielded a nominally significant p-value (p-value = 0.027), but this was no longer significant after correction for multiple testing (adjusted p-value = 0.107). Although the findings do not meet the criteria for statistical significance, they suggest a potentially meaningful trend, warranting further investigation into the A-G haplotype’s association with increased AD risk. Discussion The genetic architecture of complex diseases such as AD remains challenging to disentangle, thereby impeding the development of mechanism-based therapeutic and preventive strategies. A well-characterized example is the APOE locus, which has been associated with AD risk for over three decades. Despite extensive genetic investigations, the molecular mechanisms through which the APOE locus influences AD pathogenesis remain incompletely defined. Given the intricate pattern of genetic association signals at the APOE locus in relation to AD risk, comprehensive delineation of its haplotype architecture may offer novel insights into stratifying high- and low-risk haplotypes. Such refined haplotypic resolution could improve individual-level risk prediction and facilitate the identification of critical, pharmacologically actionable targets for disease prevention and therapeutic intervention. To advance the resolution of haplotype architecture at the APOE locus, we developed a novel haplotyping strategy based on multiplexing dPCR (mdPCR), enabling the determination of phased haplotypes from individual gDNA samples. This approach allows precise inference of allelic co-localization across multiple variants on the same chromosomal strand, thereby distinguishing maternal from paternal haplotypes with high fidelity. This procedure leverages a unique feature of the digital PCR, which isolates diploid gDNA fragments into individual microwells. Using multiplex PCR, the instrument produces a 2D plot that distinguishes gDNA fragments based on whether they carry single or double florescent signals from the two alleles. Partition counts from different quadrants of 2D plots are analyzed and categorized using GMM algorithm, revealing the coexistence of the two alleles on the same DNA fragment. This approach is scalable to incorporate more than two loci, thereby enabling resolution of haplotype architecture comprising multiple distal genetic variants. Consequently, this haplotyping tool provides a straightforward and efficient method for reconstructing precise phased haplotypes from large samples. To interrogate the APOE locus, we selected three tagging SNPs—rs2075650 ( TOMM40 ), rs429358 ( APOE ), and rs12721046 ( APOC1 )—which collectively span a 25.6 kilobase (kb) region. These SNPs exhibit strong linkage disequilibrium (LD) and have been consistently associated with AD risk, enabling targeted exploration of haplotype variation within this locus. Applying the mdPCR procedure, we resolved the haplotype configurations derived from the three selected SNPs, thereby delineating the underlying allelic architecture at the APOE locus. This distance exceeds the optimized fragment length (7 kb) typically used for long range PCR-based deep sequencing haplotype structure analysis ( 47 ). It also far exceeds the optimal read length (700-bases) of Sanger sequencing ( 48 ), a technique frequently used for haplotype reconstruction in clone-based approaches. Although nanopore sequencing can phase long haplotypes, its lower base-level accuracy—especially in homopolymers and GC-rich regions—can complicate SNP-level phasing, while its relatively modest throughput and flowcell variability may limit scalability for large cohorts. Additionally, the computational demands for long-read phasing and structural variant resolution require sophisticated algorithms and resources, making it less turnkey than short-read platforms. Unlike protocols that rely on pre-amplification of PCR fragments, which are constrained by factors such as SNP distance or the amplicon’s GC content ( 47 ), this mdPCR procedure readily circumvent these challenges. For example, modulation of sonication intensity allows precise control over the fragment length of input genomic DNA, enabling optimization of template size for downstream reactions. In cases where SNPs are surrounded by high-GC content, alternative SNPs in complete or high LD with the target can be selected as surrogate SNP candidates. When applied with a suitable anchoring SNP, mdPCR can be used to extend phased haplotypes beyond the immediate region of interest. This approach allows for the reconstruction of haplotype structures involving SNPs located hundreds of kilobases apart. With only a few targeted reactions, long-range haplotype phasing may be achieved with high accuracy While this mdPCR is a straightforward method for generating robust and precise haplotype data, we have tackled and resolved several challenges during the development of the protocol. These challenges included gDNA preparation, the instrument's capability, the specificity and compatibility of the SNP assays, and the multiplexing of PCR within the dPCR setting. These are described in Supplementary Methods. Triplex and Quadruplex Potentials and Challenges The greatest advantage of mdPCR procedure is its multiplexing capability, which reduces the number of reactions needed for phase haplotyping multiple SNPs. Beyond its cost-effectiveness, the method demands minimal manual input, enabling scalable, high-throughput processing of parallel samples. In a multiplex setting, it is possible to reconstruct haplotype of up to four target SNP sites spanning genetic regions within a single reaction on the Absolute Q system. Through the multiplexing of SNP assays, this strategy enables haplotype phasing of SNPs separated by considerable genomic distances, thereby facilitating resolution across spatially distant loci. By leveraging overlapping SNP alleles to combine adjacent haplotypes, it is possible to infer extended haplotype structures spanning hundreds of kilobases. For example, anchoring the analysis on rs429358, which defines the ε4 allele of APOE , provided a robust scaffold for interrogating allelic co-localization across distal loci on the same chromosome. This strategy substantially increased confidence in haplotype phasing and assignment. To establish proof of concept, we applied a triplex mdPCR haplotyping protocol targeting SNPs at TOMM40 (rs2075650), APOE (rs429358), and APOC1 (rs12721046), spanning over 25 kb within the genome. However, assay integration proved technically challenging, and the multiplexed design did not yield successful haplotype resolution, The reactions either failed to amplify properly, or the number of positive partitions dropped by several folds. In some cases, the rainy effect of specific SNP assay that originally was not presented in duplex reaction would appear in triplex reaction. This observation may be attributed to the increased complexity of primer-to-primer, primer-to-probe, and probe-to-probe interactions from multiple TaqMan assays. Additionally, since it is known that the ideal annealing and extension temperature of TOMM40 rs2075650 and APOC1 rs12721046 differed by 4ºC, assay incompatibility under multiplexing conditions can be anticipated due to potential cross-reactivity and design constraints. Despite this challenge, the APOE rs429358 assay can be duplexed with additional SNP assays, indicating the potential to expand the protocol to triplex or quadruplex formats, contingent on the development of suitable multiplexing conditions. Notably, despite utilizing ThermoFisher Scientific’s well-established SNP TaqMan assay design tools, none of the multiple TOMM40 rs2075650 assays we developed yielded the expected amplifications and results. This highlights the gap between the in-silico predictions and empirical performance in multiplex assays, and the need to address primer-primer, primer-probe, and probe-probe interference. High-Risk and Low-Risk Haplotypes in Alzheimer’s Diseases Assessment of pairwise phased haplotypes in relation to AD risk yielded non-significant p-values, indicating no detectable association under the tested conditions. However, we observed a potential trend suggesting that the rs2075650–rs12721046 A–G haplotype may be associated with increased AD risk. This trend is likely influenced by the limited number of control subjects in our study, which reduced statistical power. In order to validate the accuracy of our phased haplotype data, we queried the LDlink LDhap Tool database ( 49 ), and confirmed that the A–G haplotype is indeed the most common pattern observed in both APOE ε3 and ε4 carriers. Previous studies have suggested that the rs2075650–A allele is associated with human longevity ( 50 ), whereas rs2075650–g allele was linked to increased risk of developing AD ( 51 ). However, our haplotype analysis did not find a significant association between the rs2075650 g-allele and elevated AD risk among individuals carrying the APOE ε3/ε4 genotype. This null finding may, in part, reflect limitations in statistical power due to the sample size of the current cohort. Given the high LD between the SNPs rs2075650 and the SNP rs429358 (the APOE ε4 determinant SNP), there may be little to no independent effect of rs2075650 on the risk of developing AD ( 52 ). Furthermore, because our study specifically focused on phasing the haplotype structure of individuals heterozygous in APOE rs429358, any potential effect of TOMM40 SNP rs2075650 may have been diluted. It is also possible that the longevity effect of rs2075650–A allele may confound AD associations, as individuals with the A allele may live longer, and therefore be more likely to develop age-related diseases such as AD. Thus, the role of SNP rs2075650 may not be limited to a simple dichotomy of AD-risk versus AD-protective, but may reflect a more complex interplay between aging, survival, and disease manifestation. Study of Kulminski et al. reported that individuals with complete heterozygous at all three loci—rs2075650–rs429358–rs12721046 (Ag/Tc/Ga)— had a 1.59-fold higher risk of developing AD than those who did not carry any minor alleles ( 27 ). They postulated that the minor alleles of rs2075650 (g-allele) and rs12721046 (a-allele) are associated with increased AD risk. However, our findings do not align with their results. One plausible explanation for this discrepancy is the use of differing methodological approaches. Kulminski et al. used an LD contrast test to assess differences between affected and unaffected individuals, while our approach involved direct reconstructing phased haplotypes at individual level. Our results challenge the common assumption that all ‘risk’ alleles are exclusively linked to APOE ε4, thereby adding another layer of complexity to the understanding of how risk and non-risk alleles would influence gene regulations and disease manifestation. Collectively, this haplotyping result suggested a borderline significant difference in the frequency of the rs2075650-A/rs12721046-G haplotype between individuals with AD and control, suggesting a possible contribution of this haplotype to disease susceptibility. Although these findings warrant cautious interpretation due to statistical limitations, the consistent directional trend observed across multiple analyses supports a potentially meaningful association with increased AD risk. These results also highlight the necessity for replication in larger, independent cohorts and further mechanistic investigations to elucidate the functional consequences of this haplotypic configuration. Importance of Phased Haplotype in Complex Disease Studies While this pilot study did not achieve definitive stratification of ε4-associated haplotypes into distinct categories of elevated or reduced AD risk, implementation of such molecular haplotype framework substantially improved the efficiency of statistical analyses and enhanced resolution of the underlying genetic architecture. The mdPCR-based haplotyping approach offers substantial potential for enhancing the efficiency and scalability of haplotype analysis in genetically complex disorders. By precisely defining phased haplotypes for individuals, comparing these haplotypes between disease and control populations can identify haplotype structures associated with either higher or lower disease risk. Deciphering the biological effects of these haplotypes is likely to bridge the gap between genotype and phenotype. By shifting from single-variant interrogation to combinatorial allelic profiling, this strategy better captures the multidimensional complexity of genomic regulation and its potential functional consequences within the context of three-dimensional chromatin organization. For instance, haplotypes may potentially influence the coregulation of multiple genes within the topologically associated domain (TAD).( 53 ). Within these domains, chromatin interactions (CI) facilitate the physical interaction of enhancers and promoters to regulate gene expressions. Different SNPs and haplotype structure within the TAD can heavily influence regulatory efficiencies, thereby affect the phenotypic traits associated with the genes ( 54 ). Therapeutic Avenue in AD Converging studies have indicated that the three genes within the APOE locus ( TOMM40 , APOE , and APOC1 ) are co-regulated within a shared 3D chromatin architecture ( 13 , 55 – 58 ). These findings support the concept of locus-wide co-regulation with potential synergistic effects on AD-relevant pathways. This locus-wide co-regulation provides a rationale for employing mdPCR strategy to precisely resolve locus-specific haplotype configurations in cohorts. By combining mdPCR procedure with statistical phasing enables sample-level haplotypes that can be related to CI strength and haplotype-specific expression, connecting 3D genomics architecture to functional effects. As a forward-looking possibility, CI-dependent enhancers/silencers within the APOE locus could be targeted to adjust transcription in a locus-specific manner. Allele-tuned DNA mimetics can be engineered to influence chromatin looping—either recruiting CI-associated factors to promote looping or sequestering them to inhibit looping—enabling quantitative and temporal control of gene expression. After defining key elements, such constructs could be deployed singly or in multiplex to adjust expression of one or more genes across the locus, to tune TOMM40 , APOE , and APOC1 output. This therapeutic concept remains hypothesis-generating and will require rigorous evaluation of specificity, delivery, and off-target effects. Although we did not assay CI strength in our study, the observed haplotype patterns at the APOE locus point to a putative link with 3D chromatin organization, raising the prospect of locus-directed therapeutic strategies targeting enhancer or silencer activity to modulate aberrant gene expression. Future Directions on mdPCR SNPs Candidate Selection The multi-step filtering approach helps ensure that selected SNPs are not randomly chosen, but are supported by functional evidence validated by independent studies, and are therefore more likely to play a role in gene regulations and disease pathogenesis. Using FORGEdb (≥ 8) together with Regulome DB ( 1 – 5 ) also promotes consistency across databases, strengthening confidence in the prioritized SNP variants. Additionally, this methodology can serve as a framework to identify nearby surrogate SNPs for the mdPCR protocol when the index SNP’s surrounding DNA sequence is unsuitable for assay design or multiplexing. Candidate surrogates can be chosen within the same TAD and in strong LD with the index SNP (e.g., r² ≥ 0.9 in the study population), while meeting standard assay-design constraints including but not limited to amplicon size, GC contents, repetitive sequences, and secondary structures. The preserved regulatory context enables multiplex-friendly panels, and the concordance between the surrogate and index SNP should be confirmed empirically. Limitations One limitation of dPCR system is that the current recommended optimal fragment size of the sheared DNA is typically below 10 kb. The SNPs rs2075650 and rs429358 are 16.3 kb apart, which exceeds the optimal fragment size by approximately 6.3 kb. Although successful phasing was achieved in this experiment, it is unknown what is the true upper limit of the fragmented DNA that the dPCR microwells can intake. Furthermore, due to markedly different optimal annealing and extension temperatures for the SNP assays rs2075650 and rs12721046, we were unable to determine whether these two SNPs—located 25.6 kb apart—can be phased together under current assay conditions. These observations suggest that haplotype analysis may be limited to SNPs of located in relatively close proximity. SNPs that are too far apart may not be suitable to be analyzed with the current dPCR system. Another limitation of this study is that this protocol was developed on and was optimized specifically for Absolute Q. Although we conducted initial trial experimental run on QIAcuity, the newly established protocol was not fully tested on it. Additionally, this study did not evaluate the protocol on other dPCR instruments such as Digital LightCycler dPCR System (Roche), DropDx Digital PCR System (PreciGenome), Nio Digital PCR System (Stilla Technologies), or Raindrop dPCR System (BioRad). It is not known whether the optimizations developed in this study can be successfully replicated on other dPCR systems. Further studies are needed to determine if the other systems are suitable for haplotype pattern reconstruction. Conclusion Haplotype analysis is a powerful approach for delineating genetic variants associated with disease susceptibility and resilience. However, the reconstruction of phased haplotypes remains technically challenging. In this study, we developed a novel strategy leveraging digital PCR to resolve haplotypes across adjacent genomic loci. This method enabled direct single-molecule resolution of allele configurations, thereby improving interpretability of SNP interactions within the local chromatin context. Further methodological refinement will be necessary to extend phased haplotype resolution beyond biallelic configurations and encompass multi-loci architectures. By enabling precise delineation of both risk-associated and protective haplotypes, this approach establishes an efficient framework for clinical translation, supporting precision medicine and individualized therapeutic strategies. Declarations Ethics approval and consent to participate: This study used de-identified genomic DNA from human postmortem brain tissue samples. No identifiable private information was accessed, and cannot be linked to individuals. Therefore, institutional ethics approval and informed consent were not required for this study. Consent for publication: Not Applicable Availability of data and materials: All data generated or analyzed during this study are included in this published article and its supplementary information files. Competing interests: The authors declare that they have no competing interests. Funding: VA Merit Review Award Authors' contributions: S.C. wrote the main manuscript text, generated figures and tables, conducted experiments, coded for data and statistical analyses, contributed to design of the work and interpretation of data. E.L. provided suggestions on figures and manuscript, and contributed to the improvement of experiment protocol. L.L. contributed to the acquisition and isolation of the DNA samples for the experiments, and contributed to the improvement of experiment protocol. J.T. provided analytical and statistical suggestions. C.E.Y. provided major main manuscript revision, contributed to conception, design of the work and interpretation of data. All authors reviewed the manuscript. Acknowledgements: Not Applicable References Lambert JC, Ibrahim-Verbaas CA, Harold D, Naj AC, Sims R, Bellenguez C, et al. Meta-analysis of 74,046 individuals identifies 11 new susceptibility loci for Alzheimer's disease. Nat Genet. 2013;45(12):1452-8. Bellenguez C, Kucukali F, Jansen IE, Kleineidam L, Moreno-Grau S, Amin N, et al. New insights into the genetic etiology of Alzheimer's disease and related dementias. Nat Genet. 2022;54(4):412-36. AlzGene. AlzGene, Alzheimer Research Forum: Alzheimer Research Forum; 2025 [Available from: www.alzforum.org/alzgene. Kulminski AM, Jain-Washburn E, Nazarian A, Wilkins HM, Veatch O, Swerdlow RH, et al. Association of APOE alleles and polygenic profiles comprising APOE-TOMM40-APOC1 variants with Alzheimer's disease neuroimaging markers. Alzheimers Dement. 2025;21(2):e14445. Blue EE, Cheng A, Chen S, Yu CE, Alzheimer's Disease Genetics C. Association of Uncommon, Noncoding Variants in the APOE Region With Risk of Alzheimer Disease in Adults of European Ancestry. JAMA Netw Open. 2020;3(10):e2017666. Wei Y, Du X, Guo H, Han J, Liu M. Mitochondrial dysfunction and Alzheimer's disease: pathogenesis of mitochondrial transfer. Front Aging Neurosci. 2024;16:1517965. Sherman EL, Taylor RD, Go NE, Nargang FE. Effect of mutations in Tom40 on stability of the translocase of the outer mitochondrial membrane (TOM) complex, assembly of Tom40, and import of mitochondrial preproteins. J Biol Chem. 2006;281(32):22554-65. Zhu X, Perry G, Moreira PI, Aliev G, Cash AD, Hirai K, et al. Mitochondrial abnormalities and oxidative imbalance in Alzheimer disease. J Alzheimers Dis. 2006;9(2):147-53. Jong MC, Gijbels MJ, Dahlmans VE, Gorp PJ, Koopman SJ, Ponec M, et al. Hyperlipidemia and cutaneous abnormalities in transgenic mice overexpressing human apolipoprotein C1. J Clin Invest. 1998;101(1):145-52. Rouland A, Masson D, Lagrost L, Verges B, Gautier T, Bouillet B. Role of apolipoprotein C1 in lipoprotein metabolism, atherosclerosis and diabetes: a systematic review. Cardiovasc Diabetol. 2022;21(1):272. Gratchev A, Kzhyshkowska J, Kannookadan S, Ochsenreiter M, Popova A, Yu X, et al. Activation of a TGF-beta-specific multistep gene expression program in mature macrophages requires glucocorticoid-mediated surface expression of TGF-beta receptor II. J Immunol. 2008;180(10):6553-65. Zhou Q, Zhao F, Lv ZP, Zheng CG, Zheng WD, Sun L, et al. Association between APOC1 polymorphism and Alzheimer's disease: a case-control study and meta-analysis. PLoS One. 2014;9(1):e87017. Lee EG, Leong L, Chen S, Tulloch J, Yu CE. APOE Locus-Associated Mitochondrial Function and Its Implication in Alzheimer's Disease and Aging. Int J Mol Sci. 2023;24(13). Emilien G, Ponchon M, Caldas C, Isacson O, Maloteaux JM. Impact of genomics on drug discovery and clinical medicine. QJM. 2000;93(7):391-423. Slatkin M. Linkage disequilibrium--understanding the evolutionary past and mapping the medical future. Nat Rev Genet. 2008;9(6):477-85. International HapMap C. A haplotype map of the human genome. Nature. 2005;437(7063):1299-320. Tewhey R, Bansal V, Torkamani A, Topol EJ, Schork NJ. The importance of phase information for human genomics. Nat Rev Genet. 2011;12(3):215-23. Wang Y, Zhao Y, Bollas A, Wang Y, Au KF. Nanopore sequencing technology, bioinformatics and applications. Nat Biotechnol. 2021;39(11):1348-65. Burgtorf C, Kepper P, Hoehe M, Schmitt C, Reinhardt R, Lehrach H, et al. Clone-based systematic haplotyping (CSH): a procedure for physical haplotyping of whole genomes. Genome Res. 2003;13(12):2717-24. Garg S. Computational methods for chromosome-scale haplotype reconstruction. Genome Biol. 2021;22(1):101. Browning SR, Browning BL. Haplotype phasing: existing methods and new developments. Nat Rev Genet. 2011;12(10):703-14. Howie BN, Donnelly P, Marchini J. A flexible and accurate genotype imputation method for the next generation of genome-wide association studies. PLoS Genet. 2009;5(6):e1000529. Li Y, Willer CJ, Ding J, Scheet P, Abecasis GR. MaCH: using sequence and genotype data to estimate haplotypes and unobserved genotypes. Genet Epidemiol. 2010;34(8):816-34. Stephens M, Smith NJ, Donnelly P. A new statistical method for haplotype reconstruction from population data. Am J Hum Genet. 2001;68(4):978-89. Scheet P, Stephens M. A fast and flexible statistical model for large-scale population genotype data: applications to inferring missing genotypes and haplotypic phase. Am J Hum Genet. 2006;78(4):629-44. Delaneau O, Coulonges C, Zagury JF. Shape-IT: new rapid and accurate algorithm for haplotype inference. BMC Bioinformatics. 2008;9:540. Kulminski AM, Philipp I, Shu L, Culminskaya I. Definitive roles of TOMM40-APOE-APOC1 variants in the Alzheimer's risk. Neurobiol Aging. 2022;110:122-31. Ma XY, Yu JT, Wang W, Wang HF, Liu QY, Zhang W, et al. Association of TOMM40 polymorphisms with late-onset Alzheimer's disease in a Northern Han Chinese population. Neuromolecular Med. 2013;15(2):279-87. Chen YC, Chang SC, Lee YS, Ho WM, Huang YH, Wu YY, et al. TOMM40 Genetic Variants Cause Neuroinflammation in Alzheimer's Disease. Int J Mol Sci. 2023;24(4). Talwar P, Kushwaha S, Rawat C, Kaur H, Srivastava A, Agarwal R, et al. Validating a Genomic Convergence and Network Analysis Approach Using Association Analysis of Identified Candidate Genes in Alzheimer's Disease. Front Genet. 2021;12:722221. Kulminski AM, Jain-Washburn E, Philipp I, He L, Loika Y, Loiko E, et al. APOE varepsilon4 allele and TOMM40-APOC1 variants jointly contribute to survival to older ages. Aging Cell. 2022;21(12):e13730. Ward A, Crean S, Mercaldi CJ, Collins JM, Boyd D, Cook MN, et al. Prevalence of apolipoprotein E4 genotype and homozygotes (APOE e4/4) among patients diagnosed with Alzheimer's disease: a systematic review and meta-analysis. Neuroepidemiology. 2012;38(1):1-17. Liu CC, Liu CC, Kanekiyo T, Xu H, Bu G. Apolipoprotein E and Alzheimer disease: risk, mechanisms and therapy. Nat Rev Neurol. 2013;9(2):106-18. Tang MX, Stern Y, Marder K, Bell K, Gurland B, Lantigua R, et al. The APOE-epsilon4 allele and the risk of Alzheimer disease among African Americans, whites, and Hispanics. JAMA. 1998;279(10):751-5. Hendrie HC, Murrell J, Baiyewu O, Lane KA, Purnell C, Ogunniyi A, et al. APOE epsilon4 and the risk for Alzheimer disease and cognitive decline in African Americans and Yoruba. Int Psychogeriatr. 2014;26(6):977-85. Logue MW, Dasgupta S, Farrer LA. Genetics of Alzheimer's Disease in the African American Population. J Clin Med. 2023;12(16). Breeze CE, Haugen E, Gutierrez-Arcelus M, Yao X, Teschendorff A, Beck S, et al. FORGEdb: a tool for identifying candidate functional variants and uncovering target genes and mechanisms for complex diseases. Genome Biol. 2024;25(1):3. Boyle AP, Hong EL, Hariharan M, Cheng Y, Schaub MA, Kasowski M, et al. Annotation of functional variation in personal genomes using RegulomeDB. Genome Res. 2012;22(9):1790-7. Diagenode. DNA Shearing for Bioruptor Standard & Bioruptor Plus: Diagenode; n.d. [Available from: https://www.diagenode.com/files/protocols/DNA_Shearing_for_Standard_and_Plus_protocol.pdf. Van Rossum G, Python Software Foundation. Python Language Reference, Version 3.13.5. 2025. McKinney W. Data Structures for Statistical Computing in Python. In Proceedings of the 9th Python in Science Conference. 2010:51-6. Harris CR, Millman KJ, van der Walt SJ, Gommers R, Virtanen P, Cournapeau D, et al. Array programming with NumPy. Nature. 2020;585(7825):357-62. Pedregosa F, Varoquaux G, Gramfort A, Michel V, Thirion B, Grisel O, et al. Scikit-learn: Machine Learning in Python. Journal of Machine Learning Research. 2011;12(85):2825-30. Bishop CM. Pattern Recognition and Machine Learning: Springer; 2006. Reynolds DA. Gaussian Mixture Models. Encyclopedia of Biometrics: Springer; 2009. p. 659-63. Dempster AP, Laird NM, Rubin DB. Maximum likelihood from incomplete data via the EM algorithm. Journal of the Royal Statistical Society: Series B (Methodological). 1977;39(1):1-38. Brait N, Kulekci B, Goerzer I. Long range PCR-based deep sequencing for haplotype determination in mixed HCMV infections. BMC Genomics. 2022;23(1):31. Hert DG, Fredlake CP, Barron AE. Advantages and limitations of next-generation sequencing technologies: a comparison of electrophoresis and non-electrophoresis methods. Electrophoresis. 2008;29(23):4618-26. Machiela MJ, Chanock SJ. LDlink: a web-based application for exploring population-specific haplotype structure and linking correlated alleles of possible functional variants. Bioinformatics. 2015;31(21):3555-7. Chen S, Sarasua SM, Davis NJ, DeLuca JM, Boccuto L, Thielke SM, et al. TOMM40 genetic variants associated with healthy aging and longevity: a systematic review. BMC Geriatr. 2022;22(1):667. Potkin SG, Guffanti G, Lakatos A, Turner JA, Kruggel F, Fallon JH, et al. Hippocampal atrophy as a quantitative trait in a genome-wide association study identifying novel susceptibility genes for Alzheimer's disease. PLoS One. 2009;4(8):e6501. Deelen J, Beekman M, Uh HW, Helmer Q, Kuningas M, Christiansen L, et al. Genome-wide association study identifies a single major locus contributing to survival into old age; the APOE locus revisited. Aging Cell. 2011;10(4):686-98. Matharu N, Ahituv N. Minor Loops in Major Folds: Enhancer-Promoter Looping, Chromatin Restructuring, and Their Association with Transcriptional Regulation and Disease. PLoS Genet. 2015;11(12):e1005640. Liu X, Xu W, Leng F, Hao C, Kolora SRR, Li W. Prioritizing long range interactions in noncoding regions using GWAS and deletions perturbed TADs. Comput Struct Biotechnol J. 2020;18:2945-52. Bekris LM, Lutz F, Yu CE. Functional analysis of APOE locus genetic variation implicates regional enhancers in the regulation of both TOMM40 and APOE. J Hum Genet. 2012;57(1):18-25. Zhou X, Chen Y, Mok KY, Kwok TCY, Mok VCT, Guo Q, et al. Non-coding variability at the APOE locus contributes to the Alzheimer's risk. Nat Commun. 2019;10(1):3310. Haghani A, Thorwald M, Morgan TE, Finch CE. The APOE gene cluster responds to air pollution factors in mice with coordinated expression of genes that differs by age in humans. Alzheimers Dement. 2021;17(2):175-90. Nuytemans K, Lipkin Vasquez M, Wang L, Van Booven D, Griswold AJ, Rajabli F, et al. Identifying differential regulatory control of APOE varepsilon4 on African versus European haplotypes as potential therapeutic targets. Alzheimers Dement. 2022;18(10):1930-42. Additional Declarations No competing interests reported. Supplementary Files ChendPCRSupplementaryDoc.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7634101","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":521755006,"identity":"c5365bf1-b32d-4fa3-8fcf-1eade571af24","order_by":0,"name":"Sunny Chen","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAzElEQVRIiWNgGAWjYBAC+QYwZZfAwMB8gDgtBhB1yUAtbAlEaoFQB4HKeQyI1CKRe+wxD8OBPH7pnm8SPxhq5QwIuU9+Rl66MVBLseScs9skexiOGxPUwnAjx0waqCVxw43czQY8DMcSZzYQq2X/jZzHhn9I0HIwcYNEDiPQUzWJ/YR0GJx5YyY5xyA5ccaNNMPHMgYHjPkJaZFvzzGTeFNhl9g/I/nBwTcVdXJshLSAABMiRgwOE6OBgYHxB4JdR5yWUTAKRsEoGFEAAIjbP9T8onNKAAAAAElFTkSuQmCC","orcid":"","institution":"VA Puget Sound Health Care System","correspondingAuthor":true,"prefix":"","firstName":"Sunny","middleName":"","lastName":"Chen","suffix":""},{"id":521755007,"identity":"96eaa941-94be-4a54-9918-b7a2ff4f9e9f","order_by":1,"name":"Eun-Gyung Lee","email":"","orcid":"","institution":"VA Puget Sound Health Care System","correspondingAuthor":false,"prefix":"","firstName":"Eun-Gyung","middleName":"","lastName":"Lee","suffix":""},{"id":521755008,"identity":"2625fdcb-09c6-4df1-b527-837698998787","order_by":2,"name":"Lesley Leong","email":"","orcid":"","institution":"VA Puget Sound Health Care System","correspondingAuthor":false,"prefix":"","firstName":"Lesley","middleName":"","lastName":"Leong","suffix":""},{"id":521755009,"identity":"c24e3088-b54f-4bb8-8b3c-b74870df1464","order_by":3,"name":"Jessica Tulloch","email":"","orcid":"","institution":"VA Puget Sound Health Care System","correspondingAuthor":false,"prefix":"","firstName":"Jessica","middleName":"","lastName":"Tulloch","suffix":""},{"id":521755010,"identity":"8df79fde-62d7-4940-8c2d-7b9f708811e1","order_by":4,"name":"Chang-En Yu","email":"","orcid":"","institution":"VA Puget Sound Health Care System","correspondingAuthor":false,"prefix":"","firstName":"Chang-En","middleName":"","lastName":"Yu","suffix":""}],"badges":[],"createdAt":"2025-09-16 23:08:03","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7634101/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7634101/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":92493768,"identity":"0a032112-0d01-493d-8ae2-b400aff525f8","added_by":"auto","created_at":"2025-09-30 10:02:17","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":450080,"visible":true,"origin":"","legend":"","description":"","filename":"ChendPCRManuscript.docx","url":"https://assets-eu.researchsquare.com/files/rs-7634101/v1/96805a7946ef1384c2340ed7.docx"},{"id":92493770,"identity":"98c0ef41-aae4-4a6c-8a0b-92bb1bbe3014","added_by":"auto","created_at":"2025-09-30 10:02:17","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":7305,"visible":true,"origin":"","legend":"","description":"","filename":"4ff01b7d10cb4075bc3c71c2f73d70c6.json","url":"https://assets-eu.researchsquare.com/files/rs-7634101/v1/c7c9e9ee5e0c1fb8f31e8694.json"},{"id":92493771,"identity":"f3602a71-9db6-47e7-b678-82aae5db1f20","added_by":"auto","created_at":"2025-09-30 10:02:17","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":870384,"visible":true,"origin":"","legend":"","description":"","filename":"ChendPCRSupplementaryDoc.docx","url":"https://assets-eu.researchsquare.com/files/rs-7634101/v1/eab8752fb2f33834dfb5d373.docx"},{"id":92494598,"identity":"9f468e4b-a0f5-451d-9a5a-48ad7e7edf0f","added_by":"auto","created_at":"2025-09-30 10:10:17","extension":"xml","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":189770,"visible":true,"origin":"","legend":"","description":"","filename":"4ff01b7d10cb4075bc3c71c2f73d70c61enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7634101/v1/ff873d2c0c2acfb2350b33f3.xml"},{"id":92494599,"identity":"8bee02f1-86af-4f8d-9e91-90834238cfa5","added_by":"auto","created_at":"2025-09-30 10:10:17","extension":"png","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":62147,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7634101/v1/26923f992c2e4b6a4f65a6c8.png"},{"id":92493772,"identity":"3f1e48d9-156f-4f35-a011-3c2e79047859","added_by":"auto","created_at":"2025-09-30 10:02:17","extension":"jpeg","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":251076,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7634101/v1/72ca070bcc94783106a88dfb.jpeg"},{"id":92493776,"identity":"7ff397fa-e0e1-4e5e-9ac2-211b4d4b2d90","added_by":"auto","created_at":"2025-09-30 10:02:17","extension":"png","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":13860,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7634101/v1/bfc870dbddc1a1d3dde11e70.png"},{"id":92493774,"identity":"8ee184ca-aec6-4b82-9d55-145de2717c1a","added_by":"auto","created_at":"2025-09-30 10:02:17","extension":"png","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":54496,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7634101/v1/4b95845b0098e94b2040fb44.png"},{"id":92494600,"identity":"1a6f7123-ce1a-44bb-b034-9d2bbcce08e3","added_by":"auto","created_at":"2025-09-30 10:10:17","extension":"xml","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":190315,"visible":true,"origin":"","legend":"","description":"","filename":"4ff01b7d10cb4075bc3c71c2f73d70c61structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7634101/v1/61e5d406b37800db7e9b0447.xml"},{"id":92493777,"identity":"22612a1d-b107-43fc-8e5c-806bf2addee0","added_by":"auto","created_at":"2025-09-30 10:02:17","extension":"html","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":199644,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7634101/v1/7f44bf3b896fb3531db299a3.html"},{"id":92493766,"identity":"90cb1ff6-c9a4-42b4-aacf-661e3af6ab4e","added_by":"auto","created_at":"2025-09-30 10:02:16","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":101252,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eDesign and analytical pipeline for \u003c/strong\u003e\u003cem\u003e\u003cstrong\u003eAPOE\u003c/strong\u003e\u003c/em\u003e\u003cstrong\u003elocus-specific phased haplotyping. \u003c/strong\u003ea) Gene map of the \u003cem\u003eTOMM40\u003c/em\u003e–\u003cem\u003eAPOE\u003c/em\u003e–\u003cem\u003eAPOC1\u003c/em\u003e locus, highlighting the SNPs of interest—rs2075650, rs429358, and rs12721046—spanning approximately 25.6 kb.\u003cstrong\u003e \u003c/strong\u003eb) A hypothetical scenario illustrating how a heterozygous individual at each of the three SNP loci may yield up to four possible haplotype combinations for any given pair of SNPs.\u003cstrong\u003e \u003c/strong\u003ec) Using allele-specific Taqman assays targeting rs2075650-g, rs429358-c, and rs12721046-G, mdPCR enables the detection of whether these alleles reside on the same chromosome (Cis) or on different chromosomes (Trans).\u003cstrong\u003e \u003c/strong\u003ed) Based on the partitioning patterns observed in mdPCR, one can reconstruct the phased haplotypes of the sample.\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7634101/v1/f0d84776be3a929d4bd4fd5a.png"},{"id":92493767,"identity":"8c269638-1d0f-431d-b85f-9c218ff275f6","added_by":"auto","created_at":"2025-09-30 10:02:16","extension":"jpeg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":134970,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003eRepresentative 2D-plots generated by duplex dPCR system. \u003c/strong\u003eA graphical depiction of different fluorescent signals detected from 3 different samples. (a) Cis (c-G) haplotype sample that is homozygous on both SNPs, rs429358 (c/c) and rs12721046 (G/G). (b) Cis (c-G) sample that is heterozygous on both SNPs, rs429358 (T/c) and rs12721046 (G/a). (c) Trans (c-a)(T-G) sample that is heterozygous on both SNPs, rs429358 (T/c) and rs12721046 (G/a). The y-axis captures fluorescent signals emitted by FAM-probe (rs429358 c-allele), and x-axis captures fluorescent signals emitted by the TAMRA-probe (rs12721046 G-allele). The TAMRA signal is detected by ABY channel by Absolute Q dPCR system, therefore labeled as ABY Fluorescence by the software.\u003c/p\u003e","description":"","filename":"floatimage2.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7634101/v1/ff7b4d387d7b0f401f5d95af.jpeg"},{"id":92889563,"identity":"bebcebb8-6fd7-4b19-94ee-302bc4835b7f","added_by":"auto","created_at":"2025-10-06 17:31:45","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1570837,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7634101/v1/b4ffc0d1-4510-437b-8c69-873c4b415e3d.pdf"},{"id":92493769,"identity":"27feb579-ac47-4bc9-ae78-bee48e39682b","added_by":"auto","created_at":"2025-09-30 10:02:17","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":870384,"visible":true,"origin":"","legend":"","description":"","filename":"ChendPCRSupplementaryDoc.docx","url":"https://assets-eu.researchsquare.com/files/rs-7634101/v1/8f83e4ac4dd772f2c2a741a2.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"From Genotype to Phased Haplotype: Multiplex Digital PCR-Based Haplotyping at the APOE Locus in Alzheimer’s Disease","fulltext":[{"header":"Introduction","content":"\u003cp\u003eComplex disorders such as late-onset Alzheimer's disease (LOAD) are characterized by substantial genetic contributions. Large-scale genome-wide association studies (GWAS) and next-generation sequencing approaches (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e) have identified over 150 loci associated with LOAD, the majority of which confer modest risk, with reported odds ratios ranging from approximately 1.1 to 1.5. Among these, the apolipoprotein E gene (\u003cem\u003eAPOE\u003c/em\u003e) locus exhibits the most pronounced effect on disease susceptibility, with an estimated odds ratio of 3.7 (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e), markedly surpassing that of other loci. This elevated genetic risk underscores the pivotal role of \u003cem\u003eAPOE\u003c/em\u003e in modulating LOAD pathogenesis. Elucidating the genetic architecture of the \u003cem\u003eAPOE\u003c/em\u003e locus may inform more precise therapeutic targets for LOAD.\u003c/p\u003e\u003cp\u003eMultiple independent GWAS have reported robust associations between single nucleotide polymorphisms (SNPs) across three adjacent genes\u0026mdash;\u003cem\u003eTOMM40\u003c/em\u003e, \u003cem\u003eAPOE\u003c/em\u003e, and \u003cem\u003eAPOC1\u003c/em\u003e\u0026mdash;and LOAD (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e) in the \u003cem\u003eAPOE\u003c/em\u003e locus. This shared signal may reflect one of two potential underlying mechanisms: [1] the association is primarily driven by \u003cem\u003eAPOE\u003c/em\u003e, with signals from neighboring loci attributable to strong linkage disequilibrium (LD) (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e); [2] a combinatorial effect wherein variants across all three genes jointly influence disease susceptibility and related phenotypes. While distinguishing between these hypotheses remains methodologically challenging, the latter scenario introduces a compelling framework in which coordinated regulatory or functional elements across the locus exert synergistic effects on AD pathogenesis.\u003c/p\u003e\u003cp\u003eThis alternative scenario is supported by biological evidence, as characteristic AD phenotypes frequently encompass mitochondrial dysfunction and impaired innate immune responses\u0026mdash;features that may be mechanistically linked to the adjacent genes flanking \u003cem\u003eAPOE\u003c/em\u003e. Mitochondrial dysfunction is one of the hallmarks of AD (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e); reduced Tom40\u0026mdash;encoded by \u003cem\u003eTOMM40\u003c/em\u003e, a core component of the mitochondrial translocase complex\u0026mdash;compromises mitochondrial integrity (\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e), promotes metabolic dysregulation, oxidative stress, and mitochondrial damages (\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e). \u003cem\u003eAPOC1\u003c/em\u003e encodes apolipoprotein C1 involved in lipid metabolism (\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e) and innate immune responses (\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e); increased brain apoC1 level and risk-associated \u003cem\u003eAPOC1\u003c/em\u003e genetic variants have also been linked to increased AD risk (\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e). Elevated expression of \u003cem\u003eTOMM40\u003c/em\u003e, \u003cem\u003eAPOE\u003c/em\u003e, and \u003cem\u003eAPOC1\u003c/em\u003e is observed in human cellular models under oxidative stress, and in postmortem brain tissues from AD patients (\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e). To evaluate this model, it is critical to define haplotype structures associated with differential risk and to develop streamlined molecular haplotyping methodologies capable of resolving locus-level genetic architecture with high fidelity.\u003c/p\u003e\u003cp\u003eThe genetic study of SNPs has been an important tool for scientists to identify common genetic variants associated with the risk of diseases (\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e). Most disease-associated SNPs identified by GWAS are not located within genes or well-defined regulatory elements, making the functional assignment of these SNPs challenging. Through computational algorithms, studies have uncovered non-random associations of SNP alleles at different sites that are in LD (\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e). These alleles that are in LD are frequently co-inherited and can vary between different ethnic populations; the combinations of the alleles located on the same chromosome are known as haplotypes (\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eDespite recent advancements in genotyping technologies that have made SNP pattern characterization more efficient, most studies have not accounted for the phased haplotype (i.e., physically separated maternal and paternal chromosomes) of the DNA samples (\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e). Phased haplotype analysis is essential for studying complex diseases. When multiple disease-associated genetic alleles are in LD with each other, it becomes difficult to determine if these alleles collectively influence disease risk. This complexity is further compounded by whether the alleles are located on the same or different chromosomes (i.e., paternal and maternal). In a hypothetical scenario where a person inherited heterozygote major and minor allele in both SNP1 (M\u003csub\u003e1\u003c/sub\u003e/m\u003csub\u003e1\u003c/sub\u003e) and SNP2 (M\u003csub\u003e2\u003c/sub\u003e/m\u003csub\u003e2\u003c/sub\u003e) can yield 4 possible haplotype patterns: M\u003csub\u003e1\u003c/sub\u003eM\u003csub\u003e2,\u003c/sub\u003e M\u003csub\u003e1\u003c/sub\u003em\u003csub\u003e2,\u003c/sub\u003e) m\u003csub\u003e1\u003c/sub\u003eM\u003csub\u003e2\u003c/sub\u003e and m\u003csub\u003e1\u003c/sub\u003em\u003csub\u003e2\u003c/sub\u003e. The possible combinations of haplotype patterns grow exponentially as more SNP sites are added into the study. As a result, deciphering the molecular mechanisms underlying disease-associated SNPs remains challenging. This leads to phase uncertainty, which diminishes the power to accurately detect specific genetic loci or combinations of alleles at different loci that may contribute to disease risk signals. On the contrary, information gained from phased haplotype can clearly distinguish the independent or additive effects of linked allele variants. This helps to determine whether the risk of disease arises from a single locus or from a combination of multiple loci, which highlights the importance of phased haplotype analysis in complex diseases.\u003c/p\u003e\u003cp\u003eThe main challenges of phased haplotyping stem from its computational complexity and the absence of cost-effective methods (\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e). Commonly used SNP analysis techniques such as TaqMan-based genotyping and SNP microarrays do not provide sufficient information to infer the haplotype pattern when both SNP sites are heterozygous for both alleles. Although frequently used sequencing methods such as Sanger sequencing and next-generation sequencing can deliver clear sequence reads of genotype data, the information from these reads is insufficient to phase haplotypes. Although there are higher-fidelity approaches that do not rely on algorithmic inference\u0026mdash;such as nanopore sequencing, which is capable of effectively identifying haplotype structures within 25 kilobases (\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e), and clone-based systematic haplotyping, which can reconstruct haplotype of regions spanning more than 60 kilobases (\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e)\u0026mdash;, these techniques require intensive trainings, involve tedious works, are extremely expensive, and cannot process large sample sizes simultaneously. Consequently, available haplotype phasing protocols are often expensive, rely on large-scale sequencing for haplotype reconstruction, or require long and stringent process of computational assembly (\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e). The most commonly used approach applies statistical algorithms, which only estimate haplotypes with chromosome phase uncertainty and do not accurately reconstruct the original haplotypes. For example, known algorithmic methods such as IMPUTE2 (\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e), MaCH (\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e), PHASE (\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e), fastPHASE (\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e), and Shape-IT (\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e) require extremely large sample sizes for accuracy, take a long time to run, and use varying computational standards for haplotype reconstruction. Furthermore, there is no gold standard on how to perform statistical-based haplotype inference. These limitations make current techniques extremely difficult to apply into clinical setting.\u003c/p\u003e\u003cp\u003eIn this study, we have developed a new molecular haplotyping procedure to define the phased haplotypes using the digital PCR (dPCR). The goal is to offer a streamlined approach that simplifies the processing time required for haplotype reconstruction, and facilitate haplotype studies for complex diseases. Unlike existing methodologies, this haplotyping protocol is not restricted by the distance between SNP sites, is less expensive compared to sequencing, is capable of processing multiple samples simultaneously, and does not require any statistical computation to infer haplotype structure. This makes it an attractive method for conducting phased haplotype studies.\u003c/p\u003e\u003cp\u003eTo develop this haplotyping procedure, we concentrated on the haplotype structure of the human \u003cem\u003eAPOE\u003c/em\u003e gene locus, as multiple SNPs across three genes (\u003cem\u003eTOMM40\u003c/em\u003e, \u003cem\u003eAPOE\u003c/em\u003e, and \u003cem\u003eAPOC1\u003c/em\u003e) in this region have been consistently linked to LOAD by GWAS (\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e). Among these SNPs, \u003cem\u003eTOMM40\u003c/em\u003e rs2075650, \u003cem\u003eAPOE\u003c/em\u003e rs429358; and \u003cem\u003eAPOC1\u003c/em\u003e rs12721046 exhibit the strongest association signals on AD risk (\u003cspan additionalcitationids=\"CR28 CR29\" citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e), although their effects vary between ancestries due to inherited haplotype structure (31). For example, the ε4 variant (characterized by non-synonymous SNPs rs429358 and rs7412) of the gene \u003cem\u003eAPOE\u003c/em\u003e is associated with increased risk of development of AD (\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e). However, multiple studies have suggested that despite the high prevalence rate of \u003cem\u003eAPOE\u003c/em\u003e ε4 variant in African and Hispanic populations, the impact of the risk alleles on AD risk is several folds lower than Caucasian population (\u003cspan additionalcitationids=\"CR35\" citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e). The reason for such discrepancy may be due to different ancestral haplotype structures and combination of genetic alleles, which lead to diverse effects of AD risk. A potential explanation is that the risk effect of \u003cem\u003eAPOE\u003c/em\u003e ε4 allele can be further modified by adjacent genetic variants across different populational groups. Therefore, these SNPs serve as excellent examples for phased haplotype construction. By reconstructing phased haplotypes from discrete allelic variants within the \u003cem\u003eAPOE\u003c/em\u003e locus, we aim to elucidate the locus-specific genetic architecture contributing to AD susceptibility\u003c/p\u003e"},{"header":"Methodology","content":"\u003cp\u003e\u003cem\u003eDNA Extraction\u003c/em\u003e:\u003c/p\u003e\n\u003cp\u003eGenomic DNA (gDNA) were isolated from frozen post-mortem brain (PMB) tissues from the cerebellum and frontal cortex. The PMB was lysed with TissueLyser L1, 40Hz, 1 min (Qiagen). The gDNA were extracted using the AllPrep DNA/RNA Mini Kit (Qiagen). All DNA isolation procedures were performed according to the manufacturer\u0026rsquo;s protocols. Nucleic acid concentrations and qualities were assessed by NanoPhotometer (Implen), and the samples were stored at -20 \u0026deg;C prior to use.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eEvidence-Based Confirmation of SNPs Candidate\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eTo identify meaningful SNPs that may influence regulatory elements within a TAD region, we used bioinformatics tools to prioritize functional SNP variants for future phased haplotype studies. One such tool is the FORGEdb scoring system (37), which integrates multiple functional evidences. We suggest selecting SNPs with a high FORGEdb score (\u0026ge;8), as this indicates strong support from functional evidences such as activity-by-contact (ABC) interaction, chromatin looping, DNase I hotspot, expression quantitative trait locus (eQTL) association, histone mark, or overlap with transcription factor (TF) motifs.\u003c/p\u003e\n\u003cp\u003eAs an additional criterion, we recommend incorporating the RegulomeDB scoring system (38) as a secondary validation step. SNPs with a low RegulomeDB score (1\u0026ndash;5) will ensure that the result was consistent between both databases, and there was at least one evidence of DNase peak or TF binding at the SNP locations. For example, the SNPs selected in this study\u0026mdash;rs2075650 (FORGEdb score = 8, RegulomeDB score = 1b), rs429358 (FORGEdb score = 8, RegulomeDB score = 4), and rs12721046 (FORGEdb score = 8, RegulomeDB score = 5)\u0026mdash;meet these criteria.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eHaplotyping Control Templates (HCTs)\u003c/em\u003e:\u003c/p\u003e\n\u003cp\u003eEight artificially synthesized 467 base pairs double stranded chimeric DNA fragment containing genetic sequences surrounding the SNPs \u003cem\u003eTOMM40\u0026nbsp;\u003c/em\u003ers2075650, \u003cem\u003eAPOE\u0026nbsp;\u003c/em\u003ers429358, and \u003cem\u003eAPOC1\u0026nbsp;\u003c/em\u003ers12721046 (Integrated DNA Technologies). The NCBI36/hg18 genomic coordinates of the conjoined sequences included: \u003cem\u003eTOMM40\u0026nbsp;\u003c/em\u003eChr19:50087357\u0026ndash;50087503; \u003cem\u003eAPOE\u0026nbsp;\u003c/em\u003eChr19:50103719\u0026ndash;50103851; and \u003cem\u003eAPOC1\u0026nbsp;\u003c/em\u003eChr19:50113012\u0026ndash;50113198 (Supplement Figure S1).\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eSample Selection and Genotyping of TOMM40, APOE and APOC1\u003c/em\u003e:\u003c/p\u003e\n\u003cp\u003eThe gDNA isolated from PMB was used for genotyping. The \u003cem\u003eTOMM40\u003c/em\u003e SNP, rs2075650 were genotyped using the TaqMan allele discrimination assay C___3084828_20 (ThermoFisher). The \u003cem\u003eAPOE\u0026nbsp;\u003c/em\u003eSNP, rs429358 were genotyped using the TaqMan allele discrimination assay C___3084793_20 (ThermoFisher). The \u003cem\u003eAPOC1\u003c/em\u003e SNP, rs12721046 was genotyped using the TaqMan allele discrimination assay C__31478296_10 (ThermoFisher).\u003c/p\u003e\n\u003cp\u003e\u003cbr\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003edPCR Sample Preparation\u003c/em\u003e:\u003c/p\u003e\n\u003cp\u003eA total of 10 \u0026micro;l of gDNA isolated from post-mortem brain tissues (1\u0026micro;g concentration) was mixed with 90 \u0026micro;l of Tris-EDTA Buffer Solution, pH 7.4 (Sigma-Aldrich) to a final concentration of 10 ng/\u0026micro;l and a total volume of 100 \u0026micro;l. The diluted DNA sample was placed in a 0.2 \u0026micro;l Bioruptor microtubes (Diagenode) and sheared with Bioruptor UCD-200 Sonication System (Diagenode) in 4\u0026deg;C water bath for 4 cycles of 1 second on/30 seconds off sonication cycles at high settings. In the absence of validated calculation tool, we estimated the mean fragment length by scaling from Diagenode\u0026rsquo;s Bioruptor Standard benchmark, with the assumption that the fragment length is inversely proportional to total ON time. Based on the specific parameters, 8 cycles at 30 seconds on/30 seconds off (total; ON = 240 seconds) yield ~350 bp product, whereas 30 cycles (total ON = 900 seconds) yield ~200 bp products (39). We estimated the length to be \u0026asymp; , which is approximately 21\u0026ndash;45 kilobases. The sample was transferred into a 1.5 ml microcentrifuge tube (Eppendorf), and was stored in -20 \u0026deg;C. A detailed rationale for preferring sonication over restriction enzyme digestion is provided in \u003cstrong\u003eSupplementary Methods Section 1\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003edPCR Primer and Probe Design\u003c/em\u003e\u003cem\u003e:\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe Absolute Q Digital PCR System (ThermoFisher) used TaqMan-based dPCR method to determine the haplotype information of each sample. The assays were designed with the PrimerQuest Tool (Integrated DNA Technology) to ensure for the best quality. For best dPCR amplification efficiency, the expected size of the amplicon was set to be close to 100 base pairs, and the maximum size should not exceed 150 base pairs. Since Absolute Q Digital PCR System (ThermoFisher) used ROX dye as passive reference, any fluorescent dyes with excitation spectrum of 580 \u0026plusmn; 10 nm and emission spectrum of 623 \u0026plusmn; 14 nm were not be selected for probe design. Information on dPCR primers and probes is listed in Supplement Table S1. The challenges encountered and the methods used for assay optimization are described in \u003cstrong\u003eSupplementary Methods Section 2\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003edPCR Haplotyping Protocol\u003c/em\u003e:\u003c/p\u003e\n\u003cp\u003eEach dPCR reaction included 3 \u0026micro;l of sonicated gDNA input (30 ng), 2 \u0026micro;l of 5X Absolute Q DNA Digital PCR Master Mix (ThermoFisher), 1 \u0026micro;l of 10X digital PCR assay for each candidate SNPs, and add nuclease-free water up to total volume of 10 \u0026micro;l. A total of 9 \u0026micro;l of the dPCR mixture was loaded into the QuantStudio Absolute Q MAP 16 Digital PCR Plate (Applied Biosystems, ThermoFisher), and 15 \u0026micro;l of Absolute Q Isolation Buffer (Applied Biosystems, ThermoFisher) was added on top of the reaction to avoid cross-contamination and evaporation. The concentration of \u003cem\u003eAPOE\u0026nbsp;\u003c/em\u003ers429358 assay was increased to 20X due to lower primer/probe efficiency. The thermal cycling profile for dPCR genotyping of \u003cem\u003eTOMM40\u003c/em\u003e rs2075650 and \u003cem\u003eAPOE\u003c/em\u003e rs429358 was 10 min at 96\u0026ordm;C, followed by 40 cycles of 5 sec at 96\u0026ordm;C and 15 sec at 62\u0026ordm;C. The thermal cycling profile for dPCR genotyping of \u003cem\u003eAPOE\u003c/em\u003e rs429358 and \u003cem\u003eAPOC1\u003c/em\u003e rs12721046 was 10 min at 96\u0026ordm;C, followed by 40 cycles of 5 sec at 96\u0026ordm;C and 15 sec at 66\u0026ordm;C.The genotype result was analyzed with QuantStudio Absolute Q Digital PCR Software 6 (Applied Biosystems, ThermoFisher), where the fluorescence signals of the targeting SNPs were translated into partitions on a 2D-plot.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003ePhase Haplotype and Statistical Analyses\u003c/em\u003e:\u003c/p\u003e\n\u003cp\u003eThe data were analyzed using Python Version 3.13.5 (40) for Windows. Data processing and Gaussian Mixture Models (GMM) analyses were performed using built-in and open-source libraries, including pandas (41), NumPy (42), and scikit-learn (43). The haplotype data were analyzed using R-Program Version R-4.5.1. The statistical package DescTools (CRAN\u0026mdash;Package DescTools (r-project.org)) was used to perform G-Test as a likelihood ratio test to differentiate between high-risk and low-risk haplotype patterns. Standardized residuals from Chi-Square were used to determine the haplotype that has the largest deviation. Post-hoc 2x2 Fisher tests were used to determine whether or not a specific haplotype statistical significantly contribute to the difference between AD and CTRL. \u003cstrong\u003e\u003cbr\u003e\u003c/strong\u003e\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cspan\u003e\u003cstrong\u003e1. Development of a Multiplexing dPCR (mdPCR) Procedure for Molecular Haplotyping\u003c/strong\u003e\u003c/span\u003e\u003c/p\u003e\n\u003cp\u003e\u003cspan\u003e\u003cspan type=\"ItalicUnderline\" class=\"ItalicUnderline\" name=\"Emphasis\"\u003e1.1. Selection of SNPs and Design of SNP Assays\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n\u003cp\u003eWe initially selected three candidate tagging SNPs within the \u003cem\u003eAPOE\u003c/em\u003e locus\u0026mdash;each previously demonstrated to be robustly associated with LOAD risk and prioritized using bioinformatic functional annotations (FORGEdb, RegulomeDB; see Methods, \u003cem\u003eEvidence-based confirmation of candidate SNPs\u003c/em\u003e)\u0026mdash;to develop mdPCR procedure. These SNPs span three genes (\u003cem\u003eTOMM40\u003c/em\u003e [rs2075650], \u003cem\u003eAPOE\u003c/em\u003e [rs429358], and \u003cem\u003eAPOC1\u003c/em\u003e [rs12721046]) and cover a 25.6 kb genomic region (see Fig. \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003ea). We then designed TaqMan-based SNP assays (rs2075650 g-allele, rs429358 c-allele, and rs12721046 G-allele) using PrimerQuest Tool (Integrated DNA Technology). Each probe was tagged with various reporter dyes. To streamline our research and development process, we also designed and synthesized artificial chimeric DNA templates, referred to as Haplotype Control Templates (HCT). These HCT artificially combine three tagging SNPs and their flanking sequences into a condensed 467 bp DNA template. Each HCT was designed to contain a unique allele of the three SNPs (see Figure \u003cspan class=\"InternalRef\"\u003eS1\u003c/span\u003e). Consequently, a total of eight HCTs were synthesized. Each HCT was diluted and subjected to copy number quantification, permitting the deposition of individual HCT molecule into distinct dPCR partitions. It was then used in subsequent multiplexing dPCR reactions to test compatibility of assays and signals as well as validate the instruments\u0026rsquo; accuracies in detecting copy numbers.\u003c/p\u003e\n\u003cp\u003e\u003cspan type=\"ItalicUnderline\" class=\"ItalicUnderline\" name=\"Emphasis\"\u003e1.2. Testing the Compatibilities of SNP Assays, Signals, and dPCR Instruments using synthetic HCT DNA constructs\u003c/span\u003e\u003c/p\u003e\n\u003cp\u003eTo determine the most suitable instrument for the mdPCR procedure, we conducted an initial test to compare the signal sensitivity and multiplex capability of the two digital PCR systems: the QIAcuity dPCR System (Qiagen) (hereinafter referred to as \u0026ldquo;QIAcuity\u0026rdquo;) and the QuantStudio Absolute Q dPCR System (ThermoFisher) (hereinafter referred to as \u0026ldquo;Absolute Q\u0026rdquo;). To make use of the full spectrum of commercially available fluorescent dyes, we evaluated signals from various fluorescent probe combinations spanning the lowest to highest emission maxima\u0026mdash;FAM, VIC, HEX, ABY, TAMRA, ROX, Texas Red, and Cy5\u0026mdash;covering approximately 520\u0026ndash;670 nanometers (nm) (emission maxima; FAM\u0026thinsp;~\u0026thinsp;520 nm to Cy5\u0026thinsp;~\u0026thinsp;670 nm). We observed that the Absolute Q instrument consistently produced better and reproducible results. While we do not observe spectral overlaps between the dyes, the fluorescent signal from Cy5 (crimson channel; emission maxima\u0026thinsp;~\u0026thinsp;667\u0026ndash;670 nm) was consistently lower than dyes from the lower excitation/emission spectra. Therefore, when multiplexing SNP assays, the assay with lowest PCR efficiency should avoid far-red labeling, particularly with Cy5 or adjacent wavelengths. The most effective duplex combinations of fluorescent probes were FAM with VIC or HEX, FAM with ABY or TAMRA, and VIC or HEX with ABY or TAMRA. Detailed evaluations of dPCR platform capabilities are provided in \u003cstrong\u003eSupplementary Methods Section 3\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003e\u003cspan type=\"ItalicUnderline\" class=\"ItalicUnderline\" name=\"Emphasis\"\u003e1.3. Interpretation of Signals and Partitions from Output of dPCR Instrument\u003c/span\u003e\u003c/p\u003e\n\u003cp\u003eUpon completion of dPCR reaction, the Absolute Q software automatically converted the fluorescent signals from partition (or microwell) counts from each reaction into 2D plots. The partitions were separated into 4 distinct quadrants based on the fluorescent signals detected from each partition in the dPCR plate. A representative duplex reaction using FAM and VIC probes is shown in Supplement Figure S2. The upper left corner of the 2D plot is the (+,\u0026ndash;) quadrant, which tallied the partitions that emitted only the FAM fluorescent signal. The bottom right corner is the (\u0026ndash;,+) quadrant that counted the partitions that emitted only VIC fluorescent signal. The upper right corner is the (+,+) quadrant that totaled the partitions that emitted both FAM and VIC signals. The final (\u0026ndash;,\u0026ndash;) quadrant located at the bottom left corner of the plot displayed partitions that did not emit any fluorescent signals.\u003c/p\u003e\n\u003cp\u003eAt optimal dilution, the concentration of DNA is sufficiently low such that each partition within the dPCR assay is statistically occupied by no more than a single template molecule. In theory, if the specific alleles of both target SNPs are located on the same DNA template, both fluorescent signals will be detected from the same partition (see Figure S2). The (+,+) quadrant will show elevated positive fluorescent signals (see Figure S2a). Conversely, if the specific alleles of the target SNPs are located on separate DNA templates, only a single fluorescent signal will be emitted from one partition at a time. This will lead to an increase in positive partitions in the (+,\u0026ndash;) and (\u0026ndash;,+) quadrants (see Figure S2b). Thus, this mdPCR method enables the accurate determination of haplotype structures. Following successful optimization of the protocol using synthetic HCT DNA constructs, the procedure was subsequently implemented for genomic DNA analysis.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e2. Validation of mdPCR in genomic DNA Haplotyping\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cspan\u003e\u003c/span\u003e\u003c/p\u003e\n\u003cp\u003e\u003cspan type=\"ItalicUnderline\" class=\"ItalicUnderline\" name=\"Emphasis\"\u003e2.1. Selection and Preparation of Genomic DNA\u003c/span\u003e\u003c/p\u003e\n\u003cp\u003eWe first genotyped and selected a panel of PMB samples that were heterozygous at the loci of interest but could not be phased using standard TaqMan genotyping procedure. Because all sites were heterozygous, the genotype data lacked phase information, precluding unambiguous inference of haplotypes (allele co-segregation across loci). Table \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e provides the demographic information of the selected PMB samples in this study. Given the substantial molecular mass and volumetric complexity of human genomic DNA, fragmentation into smaller units is necessary to facilitate stochastic partitioning within the digital PCR array. In the standard dPCR protocol, the manufacturer recommended users to reduce the size of the input gDNA by digesting them with 1 Unit of 6-base cutting restriction enzymes. Based on this recommendation, we tested various restriction enzymes that do not cleave our targeted region to digest gDNAs prior to mdPCR. However, this approach consistently resulted in a low number of positively signaled partitions, making it difficult to differentiate between quadrant-related partition counts. To address this technical limitation, we implemented an alternative approach utilizing sonication to achieve controlled fragmentation of gDNA.\u003c/p\u003e\n\u003ctable id=\"Tab1\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\n \u003cdiv class=\"CaptionContent\"\u003e\n \u003cp\u003eSample Demographic\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eSubjects\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eAD\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eCTRL\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSample number_n\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e40\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e9\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSex-Female_n (%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e23 (57.50%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e4 (44.44%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAge at death_mean (SD)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e84.05 (9.15)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e85.78 (5.93)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAge at onset_mean (SD)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e74.95 (9.38)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eDisease Duration_mean (SD)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e10.15 (4.73)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003ePostmortem interval_mean hour (SD)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e4.99 (1.67)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5.72 (3.34)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eCERAD Score_n\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAbsent\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSparse\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eModerate\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eFrequent\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e37\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBRAAK Stage_n\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eII\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eIII\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e8\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eIV\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eV\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eVI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e31\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u0026ndash;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eUsing Diagenode sonicator to optimize the sonication protocol, we prepared multiple aliquots of the same gDNA sample, and subjected them to sonication under varying energy conditions. Initially, we performed 4 cycles of sonication protocol with 5 different energy levels (1 to 5 seconds on/30 seconds off). We compared the number of positive partitions of the signals produced by TaqMan FAM assay for \u003cem\u003eAPOE\u003c/em\u003e SNP rs429358 across samples treated at each sonication energy level. The results indicated that sonication energy at 1 second on/30 seconds off produced the best outcome. After defining the optimal energy, we then tested for the optimal number of cycles achieve best dPCR result. Using the same setting, we compared the dPCR outcomes from samples subjected to 2, 4, 6, and 8 sonication cycles. The results showed that the highest partition count was achieved with the 4 cycles protocol. Based on this parameter, the average gDNA fragment size was estimated to range between 21 and 45 kilobases. The detailed calculation methodology is provided in the \u0026ldquo;dPCR Sample Preparation\u0026rdquo; subsection of the Materials and Methods.\u003c/p\u003e\n\u003cp\u003e\u003cspan type=\"ItalicUnderline\" class=\"ItalicUnderline\" name=\"Emphasis\"\u003e2.2. Haplotyping SNPs rs2075650 (TOMM40) rs429358 (APOE) and rs12721046 (APOC1) with mdPCR\u003c/span\u003e\u003c/p\u003e\n\u003cp\u003eA panel of PMB samples carrying the \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;3/\u0026epsilon;4 genotype was selected and genotyped using commercial TaqMan allelic discrimination assays to determine allelic composition at SNPs rs2075650 and rs12721046. Samples exhibiting homozygosity at both SNP loci were excluded from subsequent molecular haplotyping, as their haplotypes could be confidently inferred without the need for further analysis. Among the remaining samples, we identified a subset that were heterozygous at rs2075650, rs12721046, or both loci. For these heterozygous samples\u0026mdash;where phase could not be inferred in silico \u0026mdash;we employed mdPCR to determine the cis- or trans-configurations of the alleles coexisting at the single-haplotype level.\u003c/p\u003e\n\u003cp\u003eBased on genomic coordinates, we selected \u003cem\u003eAPOE\u003c/em\u003e SNP rs429358 as the anchoring SNP. For each selected PMB sample heterozygous at rs2075650, rs12721046, or both loci, we conducted either one or two mdPCR reactions: one reaction for rs2075650-g-allele and rs429358-c-allele, and another for rs429358-c-allele and rs12721046-G-allele (Fig. \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003eb). Optimized mdPCR output profiles are depicted in Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e\n\u003cp\u003eFigure \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003ea shows the 2D plot of a DNA sample that is homozygous at both target sites (rs429358 c/c and rs12721046 G/G), representing two copies of the c\u0026ndash;G haplotype. This sample was used as an internal control for the mdPCR protocol to improve the clustering accuracy of the subsequent Gaussian Mixture Models (GMM) analysis (see next section). Figure \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003eb depicts the 2D plot of a sample that is heterozygous on both SNPs, rs429358 (T/c) and rs12721046 (G/a), carrying a single copy of the c\u0026ndash;G haplotype. The plot shows positive but reduced signals in the (+,+) quadrant, indicating the presence of a single c-G Cis haplotype. In contrast, the sample in Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003ec exhibits a distinct 2D plot pattern of a DNA sample that is heterozygous on both SNPs, with minimal signal in the (+,+) quadrant. This suggests the absence of a c-G haplotype, supporting a Trans haplotype structure\u0026mdash;specifically, c-a and T-G.\u003c/p\u003e\n\u003cp\u003eWe utilized the mdPCR technique and performed three independent runs for each PMB sample, generating three experimental replicates per sample. The raw partition numbers from each quadrant of the respective 2D plots were extracted and are reported in Supplement Table S2. While partition counts within the (+,+) quadrant offer a preliminary estimate of c-G haplotype copy number (e.g., 0, 1, or 2), data from three independent experimental runs revealed substantial variability, limiting the reliability of this direct approach of copy number calling.\u003c/p\u003e\n\u003cp\u003e\u003cspan type=\"ItalicUnderline\" class=\"ItalicUnderline\" name=\"Emphasis\"\u003e2.3. Interpreting mdPCR Data\u003c/span\u003e\u003c/p\u003e\n\u003cp\u003eTo improve confidence in copy number inference, we employed more robust statistical algorithms, Gaussian Mixture Models (GMM) (\u003cspan class=\"CitationRef\"\u003e44\u003c/span\u003e, \u003cspan class=\"CitationRef\"\u003e45\u003c/span\u003e), which perform pattern recognition analysis and capable of accommodating assay noise and partitioning uncertainty. This approach streamlines the interpretation of mdPCR outputs and eliminates the need for manual inspection of 2D plot signals. The GMM assumes that the data are generated from a mixture of Gaussian distributions (i.e., multiple clusters), and uses the Expectation-Maximization (EM) algorithm (\u003cspan class=\"CitationRef\"\u003e46\u003c/span\u003e) to estimate the posterior probabilities that each data point belongs to a specific cluster.\u003c/p\u003e\n\u003cp\u003eThe results from the three experimental replicates were combined to account for plate-to-plate variation. We extracted all partition numbers from the (+,+), (+,\u0026ndash;) and (\u0026ndash;,+) quadrants to calculate for the fractions for the Cis and Trans haplotypes of each sample. The \u003cstrong\u003eCis Fraction\u003c/strong\u003e was calculated as \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\varvec{C}\\varvec{i}\\varvec{s}\\:\\varvec{F}\\varvec{r}\\varvec{a}\\varvec{c}\\varvec{t}\\varvec{i}\\varvec{o}\\varvec{n}=\\:\\frac{{N}_{(+,+)}}{{N}_{(+,-)}+{N}_{(-,+)}+{N}_{(+,+)}\\:}\\)\u003c/span\u003e\u003c/span\u003e, and the \u003cstrong\u003eTrans Fraction\u003c/strong\u003e as \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\varvec{T}\\varvec{r}\\varvec{a}\\varvec{n}\\varvec{s}\\:\\varvec{F}\\varvec{r}\\varvec{a}\\varvec{c}\\varvec{t}\\varvec{i}\\varvec{o}\\varvec{n}=\\:\\frac{{N}_{(+,-)}+{N}_{(-,+)}}{{N}_{(+,-)}+{N}_{(-,+)}+{N}_{(+,+)}\\:}\\)\u003c/span\u003e\u003c/span\u003e. These fractions were then used to estimate the posterior probabilities for assigning each sample to the appropriate Cis or Trans category. Since input DNA and assay concentrations were controlled, we exclude the (\u0026ndash;,\u0026ndash;) quadrant\u0026mdash;which contains negative or non-informative partitions\u0026mdash;from the analysis. The GMM analyses for \u003cem\u003eTOMM40\u003c/em\u003e\u0026ndash;\u003cem\u003eAPOE\u003c/em\u003e haplotype (rs2075650\u0026ndash;rs429358) and \u003cem\u003eAPOE\u003c/em\u003e\u0026ndash;\u003cem\u003eAPOC1\u003c/em\u003e haplotype (rs429358\u0026ndash;rs12721046) were presented in Supplement Table S3. This approach enabled more efficient and accurate haplotype resolution, enhancing confidence in variant assignment.\u003c/p\u003e\u003cp\u003e\u003cspan type=\"ItalicUnderline\" class=\"ItalicUnderline\" name=\"Emphasis\"\u003e2.4. Reconstruction and Identification of Risk Haplotype Structure\u003c/span\u003e\u003c/p\u003e\u003cp\u003eFor samples classified as Cis-configuration for the \u003cem\u003eTOMM40\u003c/em\u003e\u0026ndash;\u003cem\u003eAPOE\u003c/em\u003e haplotype (rs2075650\u0026ndash;rs429358), the deduced haplotype is g\u0026ndash;c; whereas samples classified as Trans-configuration have the A\u0026ndash;c haplotype. For samples evaluated as Cis for the \u003cem\u003eAPOE\u003c/em\u003e\u0026ndash;\u003cem\u003eAPOC1\u003c/em\u003e haplotype (rs429358\u0026ndash;rs12721046), the inferred haplotype is c\u0026ndash;G; whereas Trans samples have the c\u0026ndash;a haplotype. Integrating both sets of pairwise haplotype data from Supplement Table S3 enabled reconstruction of the \u003cem\u003eTOMM40\u003c/em\u003e-\u003cem\u003eAPOE\u003c/em\u003e-\u003cem\u003eAPOC1\u003c/em\u003e haplotype architecture for each sample (Fig. \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003ed). The corresponding diplotypes and haplotypic configurations are detailed in Table \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e\u003ctable id=\"Tab2\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eHaplotype Patterns in \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;3/\u0026epsilon;4 PMB Samples\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\"\u003e\u003cp\u003eSampleID\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\"\u003e\u003cp\u003ers2075650\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\"\u003e\u003cp\u003ers425358\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\"\u003e\u003cp\u003ers12721046\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\"\u003e\u003cp\u003eT40-\u003c/p\u003e\u003cp\u003eAPOE\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\"\u003e\u003cp\u003eAPOE-\u003c/p\u003e\u003cp\u003eAPOC1\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\"\u003e\u003cp\u003eDiplotype\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\"\u003e\u003cp\u003e\u0026epsilon;3\u003c/p\u003e\u003cp\u003eHaplotype\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\"\u003e\u003cp\u003e\u0026epsilon;4\u003c/p\u003e\u003cp\u003eHaplotype\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\"\u003e\u003cp\u003eT40-APOC1\u003c/p\u003e\u003cp\u003eHaplotype 1\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\"\u003e\u003cp\u003eT40-APOC1\u003c/p\u003e\u003cp\u003eHaplotype 2\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTRANS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTRANS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTG|Aca\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAca\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAa\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTRANS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATG|Aca\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAca\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAa\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCIS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAta|AcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAta\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAG\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATG|AcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAG\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD5\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTRANS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTRANS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTG|Aca\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAca\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAa\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD6\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATG|AcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAG\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD7\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATG|AcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAG\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD8\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTRANS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTRANS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTG|Aca\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAca\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAa\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD9\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATG|AcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAG\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD10\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTRANS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCIS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTa|AcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003ega\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAG\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD11\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCIS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCIS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATa|gcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egG\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD12\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTRANS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTRANS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTG|Aca\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAca\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAa\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD13\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCIS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCIS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATa|gcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egG\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD14\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCIS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCIS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATa|gcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egG\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD15\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCIS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCIS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATa|gcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egG\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD16\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCIS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCIS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATa|gcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egG\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD17\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATG|AcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAG\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD18\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCIS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCIS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATa|gcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egG\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD19\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTRANS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCIS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTa|AcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003ega\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAG\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD20\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTRANS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCIS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTa|AcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003ega\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAG\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD21\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTRANS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTRANS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTG|Aca\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAca\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAa\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD22\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCIS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCIS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATa|gcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egG\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD23\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTRANS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTRANS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTG|Aca\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAca\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAa\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD24\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCIS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATa|AcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAG\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD25\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCIS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCIS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATa|gcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egG\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD26\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATG|AcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAG\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD27\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCIS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCIS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATa|gcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egG\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD28\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTRANS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTRANS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTG|Aca\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAca\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAa\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD29\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTRANS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTRANS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTG|Aca\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAca\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAa\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD30\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATG|AcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAG\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD31\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCIS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTRANS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATG|gca\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egca\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003ega\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD32\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTRANS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTG|AcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAG\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD33\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATG|AcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAG\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD34\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCIS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATG|gcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egG\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD35\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTRANS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTG|AcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAG\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD36\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTRANS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTG|AcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAG\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD37\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCIS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATG|gcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egG\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD38\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATG|AcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAG\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD39\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATG|AcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAG\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAD40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTRANS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTRANS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTG|Aca\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAca\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAa\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCTRL1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTRANS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTRANS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTG|Aca\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAca\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAa\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCTRL2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCIS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCIS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATa|gcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egG\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCTRL3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTRANS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTRANS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTG|Aca\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAca\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAa\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCTRL4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATG|AcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAG\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCTRL5\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCIS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCIS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATa|gcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egcG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egG\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCTRL6\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eaa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCIS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003e-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATa|gca\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eATa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egca\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003ega\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCTRL7\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTRANS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTRANS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTG|Aca\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAca\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAa\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCTRL8\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTRANS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTRANS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTG|Aca\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAca\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAa\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eCTRL9\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAg\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTc\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eGa\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTRANS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eTRANS\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTG|Aca\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egTG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAca\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003egG\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\"\u003e\u003cp\u003eAa\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/table\u003e\u003cp\u003e*** \u003cstrong\u003eINSERT\u003c/strong\u003e Table \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e \u003cstrong\u003eHERE\u003c/strong\u003e ***\u003c/p\u003e\u003cp\u003eFollowing reconstruction of the precise molecular haplotype structures of each sample, we subsequently proceeded to evaluate the contribution of these haplotypes and diplotype configurations to AD risk. The initial G-Test revealed no statistically significant differences between the diplotypes in AD and Control groups (G\u0026thinsp;=\u0026thinsp;11.36, p-value\u0026thinsp;=\u0026thinsp;0.25). Stratifying the samples by \u003cem\u003eAPOE\u003c/em\u003e \u0026epsilon;3 and \u0026epsilon;4 status suggested a possible trend among \u0026epsilon;4 carriers, although these results did not reach statistical significance (\u0026epsilon;3: G\u0026thinsp;=\u0026thinsp;4.24, p-value\u0026thinsp;=\u0026thinsp;0.24; \u0026epsilon;4: G\u0026thinsp;=\u0026thinsp;5.87, p-value\u0026thinsp;=\u0026thinsp;0.12). To mitigate potential type II error due to limited sample size, we focused our downstream analysis on the most informative variants\u0026mdash;\u003cem\u003eTOMM40\u003c/em\u003e rs2075650 and \u003cem\u003eAPOC1\u003c/em\u003e rs12721046\u0026mdash;and combined the data accordingly. This approach revealed a borderline significant difference between the AD and Control groups for the A-G haplotype (G\u0026thinsp;=\u0026thinsp;6.50, p-value\u0026thinsp;=\u0026thinsp;0.09).\u003c/p\u003e\u003cp\u003eTo further explore this signal, we conducted a Chi-square analysis of standardized residuals, which confirmed that the A-G haplotype showed the largest deviation. However, the result remained below the threshold for statistical significance (|Standardized Residual of A-G| = 1.70). A post hoc Fisher\u0026rsquo;s exact test yielded a nominally significant p-value (p-value\u0026thinsp;=\u0026thinsp;0.027), but this was no longer significant after correction for multiple testing (adjusted p-value\u0026thinsp;=\u0026thinsp;0.107). Although the findings do not meet the criteria for statistical significance, they suggest a potentially meaningful trend, warranting further investigation into the A-G haplotype\u0026rsquo;s association with increased AD risk.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThe genetic architecture of complex diseases such as AD remains challenging to disentangle, thereby impeding the development of mechanism-based therapeutic and preventive strategies. A well-characterized example is the \u003cem\u003eAPOE\u003c/em\u003e locus, which has been associated with AD risk for over three decades. Despite extensive genetic investigations, the molecular mechanisms through which the \u003cem\u003eAPOE\u003c/em\u003e locus influences AD pathogenesis remain incompletely defined. Given the intricate pattern of genetic association signals at the \u003cem\u003eAPOE\u003c/em\u003e locus in relation to AD risk, comprehensive delineation of its haplotype architecture may offer novel insights into stratifying high- and low-risk haplotypes. Such refined haplotypic resolution could improve individual-level risk prediction and facilitate the identification of critical, pharmacologically actionable targets for disease prevention and therapeutic intervention.\u003c/p\u003e\u003cp\u003eTo advance the resolution of haplotype architecture at the \u003cem\u003eAPOE\u003c/em\u003e locus, we developed a novel haplotyping strategy based on multiplexing dPCR (mdPCR), enabling the determination of phased haplotypes from individual gDNA samples. This approach allows precise inference of allelic co-localization across multiple variants on the same chromosomal strand, thereby distinguishing maternal from paternal haplotypes with high fidelity. This procedure leverages a unique feature of the digital PCR, which isolates diploid gDNA fragments into individual microwells. Using multiplex PCR, the instrument produces a 2D plot that distinguishes gDNA fragments based on whether they carry single or double florescent signals from the two alleles. Partition counts from different quadrants of 2D plots are analyzed and categorized using GMM algorithm, revealing the coexistence of the two alleles on the same DNA fragment. This approach is scalable to incorporate more than two loci, thereby enabling resolution of haplotype architecture comprising multiple distal genetic variants. Consequently, this haplotyping tool provides a straightforward and efficient method for reconstructing precise phased haplotypes from large samples.\u003c/p\u003e\u003cp\u003eTo interrogate the \u003cem\u003eAPOE\u003c/em\u003e locus, we selected three tagging SNPs\u0026mdash;rs2075650 (\u003cem\u003eTOMM40\u003c/em\u003e), rs429358 (\u003cem\u003eAPOE\u003c/em\u003e), and rs12721046 (\u003cem\u003eAPOC1\u003c/em\u003e)\u0026mdash;which collectively span a 25.6 kilobase (kb) region. These SNPs exhibit strong linkage disequilibrium (LD) and have been consistently associated with AD risk, enabling targeted exploration of haplotype variation within this locus. Applying the mdPCR procedure, we resolved the haplotype configurations derived from the three selected SNPs, thereby delineating the underlying allelic architecture at the \u003cem\u003eAPOE\u003c/em\u003e locus. This distance exceeds the optimized fragment length (7 kb) typically used for long range PCR-based deep sequencing haplotype structure analysis (\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e). It also far exceeds the optimal read length (700-bases) of Sanger sequencing (\u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e48\u003c/span\u003e), a technique frequently used for haplotype reconstruction in clone-based approaches. Although nanopore sequencing can phase long haplotypes, its lower base-level accuracy\u0026mdash;especially in homopolymers and GC-rich regions\u0026mdash;can complicate SNP-level phasing, while its relatively modest throughput and flowcell variability may limit scalability for large cohorts. Additionally, the computational demands for long-read phasing and structural variant resolution require sophisticated algorithms and resources, making it less turnkey than short-read platforms. Unlike protocols that rely on pre-amplification of PCR fragments, which are constrained by factors such as SNP distance or the amplicon\u0026rsquo;s GC content (\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e), this mdPCR procedure readily circumvent these challenges. For example, modulation of sonication intensity allows precise control over the fragment length of input genomic DNA, enabling optimization of template size for downstream reactions. In cases where SNPs are surrounded by high-GC content, alternative SNPs in complete or high LD with the target can be selected as surrogate SNP candidates.\u003c/p\u003e\u003cp\u003eWhen applied with a suitable anchoring SNP, mdPCR can be used to extend phased haplotypes beyond the immediate region of interest. This approach allows for the reconstruction of haplotype structures involving SNPs located hundreds of kilobases apart. With only a few targeted reactions, long-range haplotype phasing may be achieved with high accuracy While this mdPCR is a straightforward method for generating robust and precise haplotype data, we have tackled and resolved several challenges during the development of the protocol. These challenges included gDNA preparation, the instrument's capability, the specificity and compatibility of the SNP assays, and the multiplexing of PCR within the dPCR setting. These are described in Supplementary Methods.\u003c/p\u003e\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\u003ch2\u003eTriplex and Quadruplex Potentials and Challenges\u003c/h2\u003e\u003cp\u003eThe greatest advantage of mdPCR procedure is its multiplexing capability, which reduces the number of reactions needed for phase haplotyping multiple SNPs. Beyond its cost-effectiveness, the method demands minimal manual input, enabling scalable, high-throughput processing of parallel samples. In a multiplex setting, it is possible to reconstruct haplotype of up to four target SNP sites spanning genetic regions within a single reaction on the Absolute Q system. Through the multiplexing of SNP assays, this strategy enables haplotype phasing of SNPs separated by considerable genomic distances, thereby facilitating resolution across spatially distant loci. By leveraging overlapping SNP alleles to combine adjacent haplotypes, it is possible to infer extended haplotype structures spanning hundreds of kilobases. For example, anchoring the analysis on rs429358, which defines the ε4 allele of \u003cem\u003eAPOE\u003c/em\u003e, provided a robust scaffold for interrogating allelic co-localization across distal loci on the same chromosome. This strategy substantially increased confidence in haplotype phasing and assignment.\u003c/p\u003e\u003cp\u003eTo establish proof of concept, we applied a triplex mdPCR haplotyping protocol targeting SNPs at \u003cem\u003eTOMM40\u003c/em\u003e (rs2075650), \u003cem\u003eAPOE\u003c/em\u003e (rs429358), and \u003cem\u003eAPOC1\u003c/em\u003e (rs12721046), spanning over 25 kb within the genome. However, assay integration proved technically challenging, and the multiplexed design did not yield successful haplotype resolution, The reactions either failed to amplify properly, or the number of positive partitions dropped by several folds. In some cases, the rainy effect of specific SNP assay that originally was not presented in duplex reaction would appear in triplex reaction. This observation may be attributed to the increased complexity of primer-to-primer, primer-to-probe, and probe-to-probe interactions from multiple TaqMan assays. Additionally, since it is known that the ideal annealing and extension temperature of \u003cem\u003eTOMM40\u003c/em\u003e rs2075650 and \u003cem\u003eAPOC1\u003c/em\u003e rs12721046 differed by 4\u0026ordm;C, assay incompatibility under multiplexing conditions can be anticipated due to potential cross-reactivity and design constraints. Despite this challenge, the \u003cem\u003eAPOE\u003c/em\u003e rs429358 assay can be duplexed with additional SNP assays, indicating the potential to expand the protocol to triplex or quadruplex formats, contingent on the development of suitable multiplexing conditions.\u003c/p\u003e\u003cp\u003eNotably, despite utilizing ThermoFisher Scientific\u0026rsquo;s well-established SNP TaqMan assay design tools, none of the multiple \u003cem\u003eTOMM40\u003c/em\u003e rs2075650 assays we developed yielded the expected amplifications and results. This highlights the gap between the in-silico predictions and empirical performance in multiplex assays, and the need to address primer-primer, primer-probe, and probe-probe interference.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\u003ch2\u003eHigh-Risk and Low-Risk Haplotypes in Alzheimer\u0026rsquo;s Diseases\u003c/h2\u003e\u003cp\u003eAssessment of pairwise phased haplotypes in relation to AD risk yielded non-significant p-values, indicating no detectable association under the tested conditions. However, we observed a potential trend suggesting that the rs2075650\u0026ndash;rs12721046 A\u0026ndash;G haplotype may be associated with increased AD risk. This trend is likely influenced by the limited number of control subjects in our study, which reduced statistical power. In order to validate the accuracy of our phased haplotype data, we queried the LDlink LDhap Tool database (\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e), and confirmed that the A\u0026ndash;G haplotype is indeed the most common pattern observed in both \u003cem\u003eAPOE\u003c/em\u003e ε3 and ε4 carriers. Previous studies have suggested that the rs2075650\u0026ndash;A allele is associated with human longevity (\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e), whereas rs2075650\u0026ndash;g allele was linked to increased risk of developing AD (\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e). However, our haplotype analysis did not find a significant association between the rs2075650 g-allele and elevated AD risk among individuals carrying the \u003cem\u003eAPOE\u003c/em\u003e ε3/ε4 genotype. This null finding may, in part, reflect limitations in statistical power due to the sample size of the current cohort.\u003c/p\u003e\u003cp\u003eGiven the high LD between the SNPs rs2075650 and the SNP rs429358 (the \u003cem\u003eAPOE\u003c/em\u003e ε4 determinant SNP), there may be little to no independent effect of rs2075650 on the risk of developing AD (\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e). Furthermore, because our study specifically focused on phasing the haplotype structure of individuals heterozygous in \u003cem\u003eAPOE\u003c/em\u003e rs429358, any potential effect of \u003cem\u003eTOMM40\u003c/em\u003e SNP rs2075650 may have been diluted. It is also possible that the longevity effect of rs2075650\u0026ndash;A allele may confound AD associations, as individuals with the A allele may live longer, and therefore be more likely to develop age-related diseases such as AD. Thus, the role of SNP rs2075650 may not be limited to a simple dichotomy of AD-risk versus AD-protective, but may reflect a more complex interplay between aging, survival, and disease manifestation.\u003c/p\u003e\u003cp\u003eStudy of Kulminski et al. reported that individuals with complete heterozygous at all three loci\u0026mdash;rs2075650\u0026ndash;rs429358\u0026ndash;rs12721046 (Ag/Tc/Ga)\u0026mdash; had a 1.59-fold higher risk of developing AD than those who did not carry any minor alleles (\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e). They postulated that the minor alleles of rs2075650 (g-allele) and rs12721046 (a-allele) are associated with increased AD risk. However, our findings do not align with their results. One plausible explanation for this discrepancy is the use of differing methodological approaches. Kulminski et al. used an LD contrast test to assess differences between affected and unaffected individuals, while our approach involved direct reconstructing phased haplotypes at individual level. Our results challenge the common assumption that all \u0026lsquo;risk\u0026rsquo; alleles are exclusively linked to \u003cem\u003eAPOE\u003c/em\u003e ε4, thereby adding another layer of complexity to the understanding of how risk and non-risk alleles would influence gene regulations and disease manifestation.\u003c/p\u003e\u003cp\u003eCollectively, this haplotyping result suggested a borderline significant difference in the frequency of the rs2075650-A/rs12721046-G haplotype between individuals with AD and control, suggesting a possible contribution of this haplotype to disease susceptibility. Although these findings warrant cautious interpretation due to statistical limitations, the consistent directional trend observed across multiple analyses supports a potentially meaningful association with increased AD risk. These results also highlight the necessity for replication in larger, independent cohorts and further mechanistic investigations to elucidate the functional consequences of this haplotypic configuration.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\u003ch2\u003eImportance of Phased Haplotype in Complex Disease Studies\u003c/h2\u003e\u003cp\u003eWhile this pilot study did not achieve definitive stratification of ε4-associated haplotypes into distinct categories of elevated or reduced AD risk, implementation of such molecular haplotype framework substantially improved the efficiency of statistical analyses and enhanced resolution of the underlying genetic architecture. The mdPCR-based haplotyping approach offers substantial potential for enhancing the efficiency and scalability of haplotype analysis in genetically complex disorders. By precisely defining phased haplotypes for individuals, comparing these haplotypes between disease and control populations can identify haplotype structures associated with either higher or lower disease risk. Deciphering the biological effects of these haplotypes is likely to bridge the gap between genotype and phenotype. By shifting from single-variant interrogation to combinatorial allelic profiling, this strategy better captures the multidimensional complexity of genomic regulation and its potential functional consequences within the context of three-dimensional chromatin organization. For instance, haplotypes may potentially influence the coregulation of multiple genes within the topologically associated domain (TAD).(\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e). Within these domains, chromatin interactions (CI) facilitate the physical interaction of enhancers and promoters to regulate gene expressions. Different SNPs and haplotype structure within the TAD can heavily influence regulatory efficiencies, thereby affect the phenotypic traits associated with the genes (\u003cspan citationid=\"CR54\" class=\"CitationRef\"\u003e54\u003c/span\u003e).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\u003ch2\u003eTherapeutic Avenue in AD\u003c/h2\u003e\u003cp\u003eConverging studies have indicated that the three genes within the \u003cem\u003eAPOE\u003c/em\u003e locus (\u003cem\u003eTOMM40\u003c/em\u003e, \u003cem\u003eAPOE\u003c/em\u003e, and \u003cem\u003eAPOC1\u003c/em\u003e) are co-regulated within a shared 3D chromatin architecture (\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e, \u003cspan additionalcitationids=\"CR56 CR57\" citationid=\"CR55\" class=\"CitationRef\"\u003e55\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR58\" class=\"CitationRef\"\u003e58\u003c/span\u003e). These findings support the concept of locus-wide co-regulation with potential synergistic effects on AD-relevant pathways.\u003c/p\u003e\u003cp\u003eThis locus-wide co-regulation provides a rationale for employing mdPCR strategy to precisely resolve locus-specific haplotype configurations in cohorts. By combining mdPCR procedure with statistical phasing enables sample-level haplotypes that can be related to CI strength and haplotype-specific expression, connecting 3D genomics architecture to functional effects. As a forward-looking possibility, CI-dependent enhancers/silencers within the \u003cem\u003eAPOE\u003c/em\u003e locus could be targeted to adjust transcription in a locus-specific manner. Allele-tuned DNA mimetics can be engineered to influence chromatin looping\u0026mdash;either recruiting CI-associated factors to promote looping or sequestering them to inhibit looping\u0026mdash;enabling quantitative and temporal control of gene expression. After defining key elements, such constructs could be deployed singly or in multiplex to adjust expression of one or more genes across the locus, to tune \u003cem\u003eTOMM40\u003c/em\u003e, \u003cem\u003eAPOE\u003c/em\u003e, and \u003cem\u003eAPOC1\u003c/em\u003e output. This therapeutic concept remains hypothesis-generating and will require rigorous evaluation of specificity, delivery, and off-target effects. Although we did not assay CI strength in our study, the observed haplotype patterns at the \u003cem\u003eAPOE\u003c/em\u003e locus point to a putative link with 3D chromatin organization, raising the prospect of locus-directed therapeutic strategies targeting enhancer or silencer activity to modulate aberrant gene expression.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e\u003ch2\u003eFuture Directions on mdPCR SNPs Candidate Selection\u003c/h2\u003e\u003cp\u003eThe multi-step filtering approach helps ensure that selected SNPs are not randomly chosen, but are supported by functional evidence validated by independent studies, and are therefore more likely to play a role in gene regulations and disease pathogenesis. Using FORGEdb (\u0026ge;\u0026thinsp;8) together with Regulome DB (\u003cspan additionalcitationids=\"CR2 CR3 CR4\" citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e) also promotes consistency across databases, strengthening confidence in the prioritized SNP variants.\u003c/p\u003e\u003cp\u003eAdditionally, this methodology can serve as a framework to identify nearby surrogate SNPs for the mdPCR protocol when the index SNP\u0026rsquo;s surrounding DNA sequence is unsuitable for assay design or multiplexing. Candidate surrogates can be chosen within the same TAD and in strong LD with the index SNP (e.g., r\u0026sup2; \u0026ge; 0.9 in the study population), while meeting standard assay-design constraints including but not limited to amplicon size, GC contents, repetitive sequences, and secondary structures. The preserved regulatory context enables multiplex-friendly panels, and the concordance between the surrogate and index SNP should be confirmed empirically.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec18\" class=\"Section2\"\u003e\u003ch2\u003eLimitations\u003c/h2\u003e\u003cp\u003eOne limitation of dPCR system is that the current recommended optimal fragment size of the sheared DNA is typically below 10 kb. The SNPs rs2075650 and rs429358 are 16.3 kb apart, which exceeds the optimal fragment size by approximately 6.3 kb. Although successful phasing was achieved in this experiment, it is unknown what is the true upper limit of the fragmented DNA that the dPCR microwells can intake. Furthermore, due to markedly different optimal annealing and extension temperatures for the SNP assays rs2075650 and rs12721046, we were unable to determine whether these two SNPs\u0026mdash;located 25.6 kb apart\u0026mdash;can be phased together under current assay conditions. These observations suggest that haplotype analysis may be limited to SNPs of located in relatively close proximity. SNPs that are too far apart may not be suitable to be analyzed with the current dPCR system.\u003c/p\u003e\u003cp\u003eAnother limitation of this study is that this protocol was developed on and was optimized specifically for Absolute Q. Although we conducted initial trial experimental run on QIAcuity, the newly established protocol was not fully tested on it. Additionally, this study did not evaluate the protocol on other dPCR instruments such as Digital LightCycler dPCR System (Roche), DropDx Digital PCR System (PreciGenome), Nio Digital PCR System (Stilla Technologies), or Raindrop dPCR System (BioRad). It is not known whether the optimizations developed in this study can be successfully replicated on other dPCR systems. Further studies are needed to determine if the other systems are suitable for haplotype pattern reconstruction.\u003c/p\u003e\u003c/div\u003e"},{"header":"Conclusion","content":"\u003cp\u003eHaplotype analysis is a powerful approach for delineating genetic variants associated with disease susceptibility and resilience. However, the reconstruction of phased haplotypes remains technically challenging. In this study, we developed a novel strategy leveraging digital PCR to resolve haplotypes across adjacent genomic loci. This method enabled direct single-molecule resolution of allele configurations, thereby improving interpretability of SNP interactions within the local chromatin context. Further methodological refinement will be necessary to extend phased haplotype resolution beyond biallelic configurations and encompass multi-loci architectures. By enabling precise delineation of both risk-associated and protective haplotypes, this approach establishes an efficient framework for clinical translation, supporting precision medicine and individualized therapeutic strategies.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cul\u003e\n \u003cli\u003e\u003cstrong\u003eEthics approval and consent to participate:\u0026nbsp;\u003c/strong\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eThis study used de-identified genomic DNA from human postmortem brain tissue samples. No identifiable private information was accessed, and cannot be linked to individuals. Therefore, institutional ethics approval and informed consent were not required for this study.\u0026nbsp;\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003e\u003cstrong\u003eConsent for publication:\u0026nbsp;\u003c/strong\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eNot Applicable\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003e\u003cstrong\u003eAvailability of data and materials:\u0026nbsp;\u003c/strong\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eAll data generated or analyzed during this study are included in this published article and its supplementary information files.\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003e\u003cstrong\u003eCompeting interests:\u003c/strong\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eThe authors declare that they have no competing interests.\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003e\u003cstrong\u003eFunding:\u003c/strong\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eVA Merit Review Award\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003e\u003cstrong\u003eAuthors\u0026apos; contributions:\u003c/strong\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eS.C. wrote the main manuscript text, generated figures and tables, conducted experiments, coded for data and statistical analyses, contributed to design of the work and interpretation of data. E.L. provided suggestions on figures and manuscript, and contributed to the improvement of experiment protocol. L.L. contributed to the acquisition and isolation of the DNA samples for the experiments, and contributed to the improvement of experiment protocol. J.T. provided analytical and statistical suggestions. C.E.Y. provided major main manuscript revision, contributed to conception, design of the work and interpretation of data. All authors reviewed the manuscript.\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003e\u003cstrong\u003eAcknowledgements:\u003c/strong\u003e\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eNot Applicable\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eLambert JC, Ibrahim-Verbaas CA, Harold D, Naj AC, Sims R, Bellenguez C, et al. Meta-analysis of 74,046 individuals identifies 11 new susceptibility loci for Alzheimer\u0026apos;s disease. Nat Genet. 2013;45(12):1452-8.\u003c/li\u003e\n\u003cli\u003eBellenguez C, Kucukali F, Jansen IE, Kleineidam L, Moreno-Grau S, Amin N, et al. New insights into the genetic etiology of Alzheimer\u0026apos;s disease and related dementias. Nat Genet. 2022;54(4):412-36.\u003c/li\u003e\n\u003cli\u003eAlzGene. AlzGene, Alzheimer Research Forum: Alzheimer Research Forum; 2025 [Available from: www.alzforum.org/alzgene.\u003c/li\u003e\n\u003cli\u003eKulminski AM, Jain-Washburn E, Nazarian A, Wilkins HM, Veatch O, Swerdlow RH, et al. Association of APOE alleles and polygenic profiles comprising APOE-TOMM40-APOC1 variants with Alzheimer\u0026apos;s disease neuroimaging markers. Alzheimers Dement. 2025;21(2):e14445.\u003c/li\u003e\n\u003cli\u003eBlue EE, Cheng A, Chen S, Yu CE, Alzheimer\u0026apos;s Disease Genetics C. Association of Uncommon, Noncoding Variants in the APOE Region With Risk of Alzheimer Disease in Adults of European Ancestry. JAMA Netw Open. 2020;3(10):e2017666.\u003c/li\u003e\n\u003cli\u003eWei Y, Du X, Guo H, Han J, Liu M. Mitochondrial dysfunction and Alzheimer\u0026apos;s disease: pathogenesis of mitochondrial transfer. Front Aging Neurosci. 2024;16:1517965.\u003c/li\u003e\n\u003cli\u003eSherman EL, Taylor RD, Go NE, Nargang FE. Effect of mutations in Tom40 on stability of the translocase of the outer mitochondrial membrane (TOM) complex, assembly of Tom40, and import of mitochondrial preproteins. J Biol Chem. 2006;281(32):22554-65.\u003c/li\u003e\n\u003cli\u003eZhu X, Perry G, Moreira PI, Aliev G, Cash AD, Hirai K, et al. Mitochondrial abnormalities and oxidative imbalance in Alzheimer disease. J Alzheimers Dis. 2006;9(2):147-53.\u003c/li\u003e\n\u003cli\u003eJong MC, Gijbels MJ, Dahlmans VE, Gorp PJ, Koopman SJ, Ponec M, et al. Hyperlipidemia and cutaneous abnormalities in transgenic mice overexpressing human apolipoprotein C1. J Clin Invest. 1998;101(1):145-52.\u003c/li\u003e\n\u003cli\u003eRouland A, Masson D, Lagrost L, Verges B, Gautier T, Bouillet B. Role of apolipoprotein C1 in lipoprotein metabolism, atherosclerosis and diabetes: a systematic review. Cardiovasc Diabetol. 2022;21(1):272.\u003c/li\u003e\n\u003cli\u003eGratchev A, Kzhyshkowska J, Kannookadan S, Ochsenreiter M, Popova A, Yu X, et al. Activation of a TGF-beta-specific multistep gene expression program in mature macrophages requires glucocorticoid-mediated surface expression of TGF-beta receptor II. J Immunol. 2008;180(10):6553-65.\u003c/li\u003e\n\u003cli\u003eZhou Q, Zhao F, Lv ZP, Zheng CG, Zheng WD, Sun L, et al. Association between APOC1 polymorphism and Alzheimer\u0026apos;s disease: a case-control study and meta-analysis. PLoS One. 2014;9(1):e87017.\u003c/li\u003e\n\u003cli\u003eLee EG, Leong L, Chen S, Tulloch J, Yu CE. APOE Locus-Associated Mitochondrial Function and Its Implication in Alzheimer\u0026apos;s Disease and Aging. Int J Mol Sci. 2023;24(13).\u003c/li\u003e\n\u003cli\u003eEmilien G, Ponchon M, Caldas C, Isacson O, Maloteaux JM. Impact of genomics on drug discovery and clinical medicine. QJM. 2000;93(7):391-423.\u003c/li\u003e\n\u003cli\u003eSlatkin M. Linkage disequilibrium--understanding the evolutionary past and mapping the medical future. Nat Rev Genet. 2008;9(6):477-85.\u003c/li\u003e\n\u003cli\u003eInternational HapMap C. A haplotype map of the human genome. Nature. 2005;437(7063):1299-320.\u003c/li\u003e\n\u003cli\u003eTewhey R, Bansal V, Torkamani A, Topol EJ, Schork NJ. The importance of phase information for human genomics. Nat Rev Genet. 2011;12(3):215-23.\u003c/li\u003e\n\u003cli\u003eWang Y, Zhao Y, Bollas A, Wang Y, Au KF. Nanopore sequencing technology, bioinformatics and applications. Nat Biotechnol. 2021;39(11):1348-65.\u003c/li\u003e\n\u003cli\u003eBurgtorf C, Kepper P, Hoehe M, Schmitt C, Reinhardt R, Lehrach H, et al. Clone-based systematic haplotyping (CSH): a procedure for physical haplotyping of whole genomes. Genome Res. 2003;13(12):2717-24.\u003c/li\u003e\n\u003cli\u003eGarg S. Computational methods for chromosome-scale haplotype reconstruction. Genome Biol. 2021;22(1):101.\u003c/li\u003e\n\u003cli\u003eBrowning SR, Browning BL. Haplotype phasing: existing methods and new developments. Nat Rev Genet. 2011;12(10):703-14.\u003c/li\u003e\n\u003cli\u003eHowie BN, Donnelly P, Marchini J. A flexible and accurate genotype imputation method for the next generation of genome-wide association studies. PLoS Genet. 2009;5(6):e1000529.\u003c/li\u003e\n\u003cli\u003eLi Y, Willer CJ, Ding J, Scheet P, Abecasis GR. MaCH: using sequence and genotype data to estimate haplotypes and unobserved genotypes. Genet Epidemiol. 2010;34(8):816-34.\u003c/li\u003e\n\u003cli\u003eStephens M, Smith NJ, Donnelly P. A new statistical method for haplotype reconstruction from population data. Am J Hum Genet. 2001;68(4):978-89.\u003c/li\u003e\n\u003cli\u003eScheet P, Stephens M. A fast and flexible statistical model for large-scale population genotype data: applications to inferring missing genotypes and haplotypic phase. Am J Hum Genet. 2006;78(4):629-44.\u003c/li\u003e\n\u003cli\u003eDelaneau O, Coulonges C, Zagury JF. Shape-IT: new rapid and accurate algorithm for haplotype inference. BMC Bioinformatics. 2008;9:540.\u003c/li\u003e\n\u003cli\u003eKulminski AM, Philipp I, Shu L, Culminskaya I. Definitive roles of TOMM40-APOE-APOC1 variants in the Alzheimer\u0026apos;s risk. Neurobiol Aging. 2022;110:122-31.\u003c/li\u003e\n\u003cli\u003eMa XY, Yu JT, Wang W, Wang HF, Liu QY, Zhang W, et al. Association of TOMM40 polymorphisms with late-onset Alzheimer\u0026apos;s disease in a Northern Han Chinese population. Neuromolecular Med. 2013;15(2):279-87.\u003c/li\u003e\n\u003cli\u003eChen YC, Chang SC, Lee YS, Ho WM, Huang YH, Wu YY, et al. TOMM40 Genetic Variants Cause Neuroinflammation in Alzheimer\u0026apos;s Disease. Int J Mol Sci. 2023;24(4).\u003c/li\u003e\n\u003cli\u003eTalwar P, Kushwaha S, Rawat C, Kaur H, Srivastava A, Agarwal R, et al. Validating a Genomic Convergence and Network Analysis Approach Using Association Analysis of Identified Candidate Genes in Alzheimer\u0026apos;s Disease. Front Genet. 2021;12:722221.\u003c/li\u003e\n\u003cli\u003eKulminski AM, Jain-Washburn E, Philipp I, He L, Loika Y, Loiko E, et al. APOE varepsilon4 allele and TOMM40-APOC1 variants jointly contribute to survival to older ages. Aging Cell. 2022;21(12):e13730.\u003c/li\u003e\n\u003cli\u003eWard A, Crean S, Mercaldi CJ, Collins JM, Boyd D, Cook MN, et al. Prevalence of apolipoprotein E4 genotype and homozygotes (APOE e4/4) among patients diagnosed with Alzheimer\u0026apos;s disease: a systematic review and meta-analysis. Neuroepidemiology. 2012;38(1):1-17.\u003c/li\u003e\n\u003cli\u003eLiu CC, Liu CC, Kanekiyo T, Xu H, Bu G. Apolipoprotein E and Alzheimer disease: risk, mechanisms and therapy. Nat Rev Neurol. 2013;9(2):106-18.\u003c/li\u003e\n\u003cli\u003eTang MX, Stern Y, Marder K, Bell K, Gurland B, Lantigua R, et al. The APOE-epsilon4 allele and the risk of Alzheimer disease among African Americans, whites, and Hispanics. JAMA. 1998;279(10):751-5.\u003c/li\u003e\n\u003cli\u003eHendrie HC, Murrell J, Baiyewu O, Lane KA, Purnell C, Ogunniyi A, et al. APOE epsilon4 and the risk for Alzheimer disease and cognitive decline in African Americans and Yoruba. Int Psychogeriatr. 2014;26(6):977-85.\u003c/li\u003e\n\u003cli\u003eLogue MW, Dasgupta S, Farrer LA. Genetics of Alzheimer\u0026apos;s Disease in the African American Population. J Clin Med. 2023;12(16).\u003c/li\u003e\n\u003cli\u003eBreeze CE, Haugen E, Gutierrez-Arcelus M, Yao X, Teschendorff A, Beck S, et al. FORGEdb: a tool for identifying candidate functional variants and uncovering target genes and mechanisms for complex diseases. Genome Biol. 2024;25(1):3.\u003c/li\u003e\n\u003cli\u003eBoyle AP, Hong EL, Hariharan M, Cheng Y, Schaub MA, Kasowski M, et al. Annotation of functional variation in personal genomes using RegulomeDB. Genome Res. 2012;22(9):1790-7.\u003c/li\u003e\n\u003cli\u003eDiagenode. DNA Shearing for Bioruptor Standard \u0026amp; Bioruptor Plus: Diagenode; n.d. [Available from: https://www.diagenode.com/files/protocols/DNA_Shearing_for_Standard_and_Plus_protocol.pdf.\u003c/li\u003e\n\u003cli\u003eVan Rossum G, Python Software Foundation. Python Language Reference, Version 3.13.5. 2025.\u003c/li\u003e\n\u003cli\u003eMcKinney W. Data Structures for Statistical Computing in Python. In Proceedings of the 9th Python in Science Conference. 2010:51-6.\u003c/li\u003e\n\u003cli\u003eHarris CR, Millman KJ, van der Walt SJ, Gommers R, Virtanen P, Cournapeau D, et al. Array programming with NumPy. Nature. 2020;585(7825):357-62.\u003c/li\u003e\n\u003cli\u003ePedregosa F, Varoquaux G, Gramfort A, Michel V, Thirion B, Grisel O, et al. Scikit-learn: Machine Learning in Python. Journal of Machine Learning Research. 2011;12(85):2825-30.\u003c/li\u003e\n\u003cli\u003eBishop CM. Pattern Recognition and Machine Learning: Springer; 2006.\u003c/li\u003e\n\u003cli\u003eReynolds DA. Gaussian Mixture Models. Encyclopedia of Biometrics: Springer; 2009. p. 659-63.\u003c/li\u003e\n\u003cli\u003eDempster AP, Laird NM, Rubin DB. Maximum likelihood from incomplete data via the EM algorithm. Journal of the Royal Statistical Society: Series B (Methodological). 1977;39(1):1-38.\u003c/li\u003e\n\u003cli\u003eBrait N, Kulekci B, Goerzer I. Long range PCR-based deep sequencing for haplotype determination in mixed HCMV infections. BMC Genomics. 2022;23(1):31.\u003c/li\u003e\n\u003cli\u003eHert DG, Fredlake CP, Barron AE. Advantages and limitations of next-generation sequencing technologies: a comparison of electrophoresis and non-electrophoresis methods. Electrophoresis. 2008;29(23):4618-26.\u003c/li\u003e\n\u003cli\u003eMachiela MJ, Chanock SJ. LDlink: a web-based application for exploring population-specific haplotype structure and linking correlated alleles of possible functional variants. Bioinformatics. 2015;31(21):3555-7.\u003c/li\u003e\n\u003cli\u003eChen S, Sarasua SM, Davis NJ, DeLuca JM, Boccuto L, Thielke SM, et al. TOMM40 genetic variants associated with healthy aging and longevity: a systematic review. BMC Geriatr. 2022;22(1):667.\u003c/li\u003e\n\u003cli\u003ePotkin SG, Guffanti G, Lakatos A, Turner JA, Kruggel F, Fallon JH, et al. Hippocampal atrophy as a quantitative trait in a genome-wide association study identifying novel susceptibility genes for Alzheimer\u0026apos;s disease. PLoS One. 2009;4(8):e6501.\u003c/li\u003e\n\u003cli\u003eDeelen J, Beekman M, Uh HW, Helmer Q, Kuningas M, Christiansen L, et al. Genome-wide association study identifies a single major locus contributing to survival into old age; the APOE locus revisited. Aging Cell. 2011;10(4):686-98.\u003c/li\u003e\n\u003cli\u003eMatharu N, Ahituv N. Minor Loops in Major Folds: Enhancer-Promoter Looping, Chromatin Restructuring, and Their Association with Transcriptional Regulation and Disease. PLoS Genet. 2015;11(12):e1005640.\u003c/li\u003e\n\u003cli\u003eLiu X, Xu W, Leng F, Hao C, Kolora SRR, Li W. Prioritizing long range interactions in noncoding regions using GWAS and deletions perturbed TADs. Comput Struct Biotechnol J. 2020;18:2945-52.\u003c/li\u003e\n\u003cli\u003eBekris LM, Lutz F, Yu CE. Functional analysis of APOE locus genetic variation implicates regional enhancers in the regulation of both TOMM40 and APOE. J Hum Genet. 2012;57(1):18-25.\u003c/li\u003e\n\u003cli\u003eZhou X, Chen Y, Mok KY, Kwok TCY, Mok VCT, Guo Q, et al. Non-coding variability at the APOE locus contributes to the Alzheimer\u0026apos;s risk. Nat Commun. 2019;10(1):3310.\u003c/li\u003e\n\u003cli\u003eHaghani A, Thorwald M, Morgan TE, Finch CE. The APOE gene cluster responds to air pollution factors in mice with coordinated expression of genes that differs by age in humans. Alzheimers Dement. 2021;17(2):175-90.\u003c/li\u003e\n\u003cli\u003eNuytemans K, Lipkin Vasquez M, Wang L, Van Booven D, Griswold AJ, Rajabli F, et al. Identifying differential regulatory control of APOE varepsilon4 on African versus European haplotypes as potential therapeutic targets. Alzheimers Dement. 2022;18(10):1930-42.\u003c/li\u003e\n\u003cli\u003e\u003c/li\u003e\n\u003cli\u003e\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"APOE, dPCR, Genotype, Phased Haplotype, Alzheimer’s Disease","lastPublishedDoi":"10.21203/rs.3.rs-7634101/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7634101/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackgrounds:\u003c/h2\u003e\u003cp\u003eAlzheimer\u0026rsquo;s disease (AD) risk reflects multi-locus variant effects. Despite advances in genotyping techniques, current molecular haplotyping approaches involve intricate operational frameworks and incur high implementation costs, and most studies therefore rely on unphased genotypes. We developed a straightforward approach to reconstruct locus-wide haplotype structure within the \u003cem\u003eAPOE\u003c/em\u003e locus and assess for potential AD-risk haplotypes.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e\u003cp\u003eWe analyzed 49 postmortem brain samples (40 AD and 9 Controls) heterozygous for \u003cem\u003eAPOE\u003c/em\u003e ε3/ε4. We developed a Multiplex Digital PCR assay to resolve allelic phase configurations of AD-associated SNPs across three closely linked loci within the \u003cem\u003eAPOE\u003c/em\u003e genomic region: \u003cem\u003eTOMM40\u003c/em\u003e (rs2075650), \u003cem\u003eAPOE\u003c/em\u003e (rs429358) and \u003cem\u003eAPOC1\u003c/em\u003e (rs12721046). Gaussian mixture modeling was used to reconstruct sample-level phased haplotype structure.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e\u003cp\u003eWe achieved high-confidence reconstruction of three-SNP haplotypes spanning a 25.6 kilobase (kb) region. The diplotype frequencies did not differ significantly in AD and Control groups (G\u0026thinsp;=\u0026thinsp;11.36, p\u0026thinsp;=\u0026thinsp;0.25). A suggestive trend was observed between the AD and Control groups for \u003cem\u003eTOMM40\u003c/em\u003e rs2075650-A and \u003cem\u003eAPOC1\u003c/em\u003e rs12721046-G haplotype (G\u0026thinsp;=\u0026thinsp;6.50, p\u0026thinsp;=\u0026thinsp;0.09).\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e\u003cp\u003eAlthough underpowered and not statistically significant, this proof-of-concept shows that multiplex digital PCR can support simple, sample-level haplotype phasing at the \u003cem\u003eAPOE\u003c/em\u003e locus. This strategy provides a robust platform for mechanistic investigation and translational application by integrating phased haplotype configurations with three-dimensional chromatin architecture\u0026ndash;associated regulatory dynamics, thereby informing locus-specific therapeutic targeting.\u003c/p\u003e","manuscriptTitle":"From Genotype to Phased Haplotype: Multiplex Digital PCR-Based Haplotyping at the APOE Locus in Alzheimer’s Disease","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-09-30 10:02:12","doi":"10.21203/rs.3.rs-7634101/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"2930e395-1b51-4f96-8bd7-ada1f39b7ed4","owner":[],"postedDate":"September 30th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-02-19T12:53:50+00:00","versionOfRecord":[],"versionCreatedAt":"2025-09-30 10:02:12","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7634101","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7634101","identity":"rs-7634101","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}