Automatic variant prioritization in suspected genetic kidney disease using the Nephro Candidate Score (N-CS)

Automatic variant prioritization in suspected genetic kidney disease using the Nephro Candidate Score (N-CS) | medRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (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];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-P4HH5NV'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search Automatic variant prioritization in suspected genetic kidney disease using the Nephro Candidate Score (N-CS) View ORCID Profile Nina Rank , Sören Lukassen , Manuel Anderegg , Kai-Uwe Eckardt , Jan Halbritter , Bernt Popp doi: https://doi.org/10.1101/2025.09.29.25336840 Nina Rank 1 Department of Nephrology and Medical Intensive Care, Charité – Universitätsmedizin Berlin , Berlin, Germany 2 Berlin Institute of Health at Charité - Universitätsmedizin Berlin, Center of Functional Genomics , Hessische Straße 4A, 10115, Berlin, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Nina Rank For correspondence: ranknina{at}gmail.com Sören Lukassen 3 Berlin Institute of Health at Charité - Universitätsmedizin Berlin, Center of Digital Health , Luisenstraße 65, 10117, Berlin, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Manuel Anderegg 1 Department of Nephrology and Medical Intensive Care, Charité – Universitätsmedizin Berlin , Berlin, Germany 4 Department of Nephrology and Hypertension, Bern University Hospital , Bern, Switzerland Find this author on Google Scholar Find this author on PubMed Search for this author on this site Kai-Uwe Eckardt 1 Department of Nephrology and Medical Intensive Care, Charité – Universitätsmedizin Berlin , Berlin, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Jan Halbritter 1 Department of Nephrology and Medical Intensive Care, Charité – Universitätsmedizin Berlin , Berlin, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Bernt Popp 2 Berlin Institute of Health at Charité - Universitätsmedizin Berlin, Center of Functional Genomics , Hessische Straße 4A, 10115, Berlin, Germany 5 Labor Berlin - Charité Vivantes GmbH , Berlin, Germany Find this author on Google Scholar Find this author on PubMed Search for this author on this site Abstract Full Text Info/History Metrics Supplementary material Data/Code Preview PDF Abstract Research Question Despite the identification of >700 genes linked to rare and inherited kidney diseases (IKD), many individuals with presumed IKD do not receive a diagnosis through genetic testing of known disease genes. Therefore, the identification of new disease genes is crucial to ending diagnostic odysseys, improving genetic counseling, and expanding treatment options. While the generation of large-scale sequencing data is no longer a substantial bottleneck, its interpretation remains challenging and offers room for improvement, notably in the discovery of novel disease genes. Methods We developed the Nephro Candidate Score (N-CS), a machine learning (ML) tool that prioritizes variants by combining a Nephro Gene Score (N-GS), a Nephro Variant Score (N-VS), and an Inheritance Score (IS). The ML-based N-GS and N-VS were trained on a wide range of genomic features to predict gene-disease relevance and variant pathogenicity, while the IS incorporates the mode of inheritance via a scoring heuristic. A Gene Set Enrichment Analysis (GSEA) was used to test whether genes top-ranked by the N-GS were enriched for kidney-related biological processes. Additionally, we tested the N-CS on an independent set of novel IKD candidate genes identified through a systematic literature search to validate its real-world performance. Results The machine learning models for the N-CS subscores demonstrated high predictive accuracy, with an XGBoost algorithm for the N-GS achieving an AUC of 0.94 and a Logistic Regression model for the N-VS reaching an AUC of 0.99 in independent test sets. The biological relevance of the N-GS ranking was confirmed by the GSEA showing a significant enrichment of kidney-associated biological processes among top-scoring genes (p < 0.001). In the independent validation using recently published literature, the N-CS assigned compellingly high scores to the majority (10 of 11) of novel candidate genes for kidney disease, demonstrating its ability to generalize to new discoveries. Conclusion The N-CS is a robust digital solution that can accelerate disease gene discovery and comes with the potential to reduce time to diagnosis. To support standardization and collaboration, the full N-CS framework is freely available, including a user-friendly web tool (NC-Scorer: https://nc-scorer.kidney-genetics.org/ ) and a command-line interface for high-throughput analysis, enabling standardized, sharable evaluation of candidate variants. Introduction Chronic kidney disease (CKD) is emerging as a global health concern with a prevalence of more than 10% in adults and is predicted to be the fifth leading cause of death worldwide by 2040 [ 1 ]. Traditionally, hypertension and diabetes mellitus are thought to represent about 2/3 of CKD causes. However, this view is mainly based on clinical assumptions rather than histological confirmation, and hypertensive nephropathy has been questioned as a stand-alone disease entity. Through the advent of modern molecular genetics as part of CKD differential diagnostics, the awareness of genetic etiologies, both monogenic and complex, has risen tremendously. It has been shown that at least 28% of patients with moderate CKD have a family history [ 2 , 3 ]. Established monogenic etiologies, such as Autosomal Dominant Polycystic Kidney Disease (ADPKD), Alport Syndrome (AS) and partially Congenital Anomalies of the Kidney and Urinary Tract (CAKUT) account for 30-50% of CKD in children and 10-30% in adults [ 4 ]. Despite the identification of >700 genes [ 5 , 6 ] being associated with rare and inherited kidney diseases, there is still a significant proportion of patients not receiving a definitive diagnosis, as no diagnostic germline variants can be found with conventional analyses (e.g. 13% in ADPKD [ 7 ], >80% in CAKUT [ 8 ]). For this reason, identification of new disease genes and novel pathogenic variants is an urgent medical need to avoid the diagnostic odysseys, improve genetic counseling, and expand treatment options. Advances in high-throughput sequencing techniques, such as exome sequencing, enable relatively inexpensive genetic testing in individuals with CKD of unexplained cause (CKDx) [ 9 ]. When diagnostic analysis of known genes is inconclusive, research-based variant filtering, evaluation, and interpretation become difficult due to large data sets. In principle, any rare variant that segregates with the clinical trait of interest may be disease-causing. But functional in vitro or in vivo investigation of all candidate variants is impossible. It is therefore necessary to prioritize and focus on the most promising variants. For this purpose, variants are often ranked manually by experts according to multiple criteria, including minor allele frequencies in large populations, inheritance mode, or the predicted effect on the encoded protein [ 10 ]. In addition, there is a variety of databases that provide information on the associated genes (e.g. on data from mouse models, protein-protein interactions, and RNA-seq data in general and specifically in the kidney). However, manually examining the various data sources for the entirety of candidate variants is extremely time-consuming, not scalable, and highly investigator-dependent. Therefore, there is a high need for an instrument that systematically and automatically prioritizes candidate variants for genetic kidney disease. Here, we present the Nephro Candidate Score (N-CS), a machine learning framework designed to address this challenge by providing a standardized, automated, and resource-efficient assessment of candidate variants. Methods Nephro Candidate Score We developed the Nephro Candidate Score (N-CS) as a composite of three subscores, the Nephro Variant Score (N-VS), the Nephro Gene Score (N-GS), and the Inheritance Score (IS), incorporating properties of the genetic variant itself, its corresponding gene, and its mode of inheritance, respectively. The N-CS ranges from 0.2 to 10 and the subscores are weighted and combined as follows: A schematic illustration of the N-CS components is depicted in Fig. 1 . Download figure Open in new tab Fig. 1. Components of the Nephro Candidate Score (N-CS). Nephro Gene Score Model development The N-GS is a ML-based model to classify genes as either associated with kidney diseases or not. The model uses two labels: Positive Genes and Negative Genes. Positive Genes are defined as those with strong evidence for a role in kidney disease, sourced from the Kidney-Genetics database [ 6 ]. This publicly accessible resource aggregates gene-disease associations from five distinct streams of evidence: expert-curated lists, comprehensive literature reviews, clinical diagnostic panels, ontology-based searches, and automated literature extraction, yielding a maximum possible evidence count (EC) of 5. We defined Positive Genes as any gene with an EC ≥ 2, reflecting a robust level of evidence supporting a gene-disease link. Negative Genes, or “dispensable” genes, include those with homozygous loss-of-function variants in gnomAD [ 11 ] (following the method from Leitao et al. [ 12 ]), no OMIM [ 13 ] phenotype, and no entry in the Kidney-Genetics database. The 109 features used in model training, as described in Table S1, include data from gnomAD, information on paralogues, promoter and exon CpG sites, promoter and exon conservation, mouse models, GTEx, and ssRNA data from the adult and fetal kidneys. We trained 12 scikit-learn [ 14 ] ML algorithms to classify genes into Positive and Negative ones as described above (Supplementary Methods, Section S1). Among the 12 trained algorithms, an XGBoost model based on a reduced set of 25 features was selected as the final algorithm, as it achieved the highest and most stable AUC. The N-GS ranges from 0 to 1, with a score of 1 indicating the highest probability that a gene is associated with kidney disease. Gene Set Enrichment Analysis To further validate the predictive performance of our model in identifying kidney disease-associated genes, we conducted a two-step Gene Set Enrichment Analysis (GSEA) [ 15 , 16 ] (Supplementary Methods, Section S2). This approach sought to determine whether genes involved in kidney-related biological processes were significantly overrepresented among the top-ranked genes predicted by our model. Nephro Variant Score The N-VS was developed as an ML based model to differentiate pathogenic variants from benign variants. To build the training and test set, we extracted all ClinVar [ 17 ] variants located in genes linked to kidney disease or considered candidate genes. Specifically, we restricted the selection to genes with an EC ≥ 2, as listed in the Kidney-Genetics database. Subsequently, we selected only the variants classified as “(Likely) Benign” and “(Likely) Pathogenic” in ClinVar to serve as labels for training the N-VS model. The initial feature set included 35 variant-level annotations derived using the Variant Effect Predictor (VEP) tool [ 18 ] (Table S2). The Supplementary Methods, Section S3, provides details of the feature preprocessing, dataset stratification, and sampling strategy used during model development. We evaluated two fundamental machine learning algorithms, decision trees and logistic regression, both of which demonstrated excellent performance. We chose logistic regression with a reduced set of 8 features as the final model because of its ability to provide continuous probability predictions when used in the N-CS. The N-VS assigns a continuous score ranging from 0 to 1, with higher values indicating a greater likelihood of a variant being pathogenic. Inheritance Score To quantitatively evaluate a variant’s mode of inheritance, we developed the Inheritance Score (IS). This score integrates this characteristic with family segregation data to create a single, informative metric. Each variant was assigned a base score reflecting its inheritance pattern (e.g., de novo, autosomal-recessive), with higher scores given to patterns with stronger pathogenic potential. When segregation evidence was available, the score was boosted toward 1.0 in proportion to the statistical strength of the segregation p-value. Conversely, for inheritance patterns where segregation evidence was expected but missing, the score was systematically reduced to reflect the increased uncertainty. The final Inheritance Score typically ranges from 0.1 (for variants with an unknown pattern) up to 0.95 for high-confidence de novo variants but can approach 1.0 in the presence of strong segregation data. A detailed description of the score derivation is provided in the Supplementary Methods, Section S4. Validation on Novel Candidate Genes from Literature To validate the N-CS’s ability to prioritize novel disease genes, we performed a systematic literature review to identify recently published candidate genes for IKD. The search was conducted using PubTator [ 19 ] with a broad query designed to capture relevant publications from 2024 and 2025 in high-impact genetics and nephrology journals (see Supplementary Methods S5 for the detailed search strategy). From this review, we identified 11 novel candidate genes that were not part of the N-CS model’s training set. We collected a total of 46 variants reported in these genes. All variants were then scored using the N-CS framework to assess whether the score could independently prioritize these new gene-disease associations (Supplementary Results S9, Table S7). Web and Command-Line Implementations of the N-CS To ensure broad accessibility and promote standardized variant assessment, we developed the Nephro Candidate Scorer (NC-Scorer), a user-friendly web tool available at https://nc-scorer.kidney-genetics.org/ . The NC-Scorer provides an intuitive interface for researchers and clinicians to apply the N-CS framework without requiring bioinformatics expertise. The tool features an interface with three primary search modes ( Fig. 2 ): Download figure Open in new tab Figure 2. The NC-Scorer web interface. The tool provides a straightforward interface for calculating the Nephro Candidate Score (N-CS). Users can input a variant, specify the inheritance pattern, and receive a score. The platform also supports dedicated gene and variant searches, as well as batch processing. Scoring Search: Users can enter a single variant in Variant Call Format (VCF), which specifies chromosome, position, and allele changes (e.g. 1-55051215-G-GA), or the transcript-based Human Genome Variation Society (HGVS) notation (e.g., NM_001009944.3:c.11935C>T). After entering the variant, users can specify the mode of inheritance to calculate a precise N-CS. The tool returns the final composite N-CS, broken down into its subscores (N-GS, N-VS, and IS) and supported by the underlying annotations. Variant Search: This allows users to look up a variant using common identifiers like HGVS notation or genomic coordinates. The result is a detailed view of the variant’s pathogenicity score (N-VS) and the gene’s relevance score (N-GS), along with all contributing evidence like CADD scores and population frequencies from gnomAD. Gene Search: Users can enter a gene symbol without further variant information to retrieve the N-GS. The main navigation bar additionally provides access to two key features: Batch scoring: To support the needs of larger research projects, the NC-Scorer includes a batch processing function that enables the simultaneous scoring of up to 200 variants at once. Results from any search can be easily exported in multiple formats, including CSV/TSV for spreadsheet analysis, JSON for programmatic use, or VCF for integration into genomic analysis pipelines. Other user-centric features include API caching for faster queries and a theme toggle for user comfort. Genes Score Overview: A comprehensive and sortable table lists all genes with their pre-computed Nephro-Gene Score (N-GS), allowing researchers to browse the entire ranked list based on kidney disease relevance. For users who require high-throughput analysis or wish to integrate the N-CS into automated workflows, a full-featured command-line interface is also available. Both the web tool and the command-line interface are open-source and freely available to the scientific community, facilitating standardized and sharable evaluation of candidate variants. Results Nephro Gene Score Performance and Distribution of the Nephro Gene Score For the N-GS, an XGBoost model with a reduced set of 25 features was chosen as the final model. It achieved an AUC of 0.943 on the independent test set (see also Supplementary Results, Section S5, Table S3). For feature importance, see Supplementary Results, Section S6, Fig. S1). Fig. 3a presents the N-GS distribution across dispensable genes, genes with an evidence count ranging from 0 to 5, and “novel genes.” Novel genes are defined as those lacking both an EC and a classification as “dispensable”. As expected, dispensable genes exhibit a very low median prediction score (0.03). Conversely, the median N-GS showed a strong positive correlation with prior evidence, increasing progressively with the EC. Notably, the novel genes displayed a wide, U-shaped distribution with a median score of approximately 0.44 ( Fig. 3b ). This bimodal pattern is consistent with the presence of two distinct subpopulations: a large group with scores clustering near zero, likely representing genes not involved in kidney disease, and a second, high-scoring group. It is within this high-scoring subset that potential new candidate genes are most likely to be found, warranting further investigation, particularly if they also harbor variants with high N-VS or IS. Download figure Open in new tab Figure 3. Distribution of Nephro Gene Scores (N-GS) across dispensable genes, genes with evidence counts (0–5), and novel genes (a) and distribution of the N-GS across novel genes (b). Novel genes do not have an evidence count, but are not classified as dispensable either. Yellow = genes used for ML training/testing, blue = genes with minor evidence for being kidney disease associated, grey = novel genes. Gene Set Enrichment Analysis The Gene Ontology (GO) terms GO:0003014 (“renal system process”) and GO:0072001 (“renal system development”), including all their descendant terms, constituted a total of 633 terms. The ‘GO_Biological_Process_2023’ dataset in the Enrichr database contained 42 of these. The latter included 5406 terms in total. Among the top 20 gene sets identified in the initial GSEA, 11 were linked to kidney-related GO terms. The Supplementary Results include a GSEA enrichment plot of the top five ranked terms, as well as a table with the top 20 ranked terms and their associated statistics (Fig. S2, Table S4). The subsequent enrichment analysis assessed whether the highest-ranked GO terms from the initial GSEA were specifically enriched for kidney-related biological processes. The kidney-GO-term set showed a highly significant enrichment, with a normalized enrichment score (NES) of 4.4 and a p-value of < 0.001. Nephro Variant Score As the final model for the N-VS, we chose a logistic regression model with a reduced set of 8 features that achieved an overall Area Under the Curve (AUC) of 0.991 on the independent test set (for further performance metrics, see Supplementary Results, Section S8, Table S5). SHAP analysis identified the most influential features in descending order of importance: the variant’s predicted functional impact, CADD-PHRED deleteriousness score, classification as a missense variant, allele frequency (AF) in gnomAD exomes, classification as a synonymous variant, classification as a frameshift variant, AF in gnomAD genomes, and finally classification as a stop-gained variant (for feature descriptions, see Table S1). Including additional features led to only a marginal increase in AUC. Additionally, we evaluated the AUC across different functional impact groups, with the results presented in Table S6. The N-CS Prioritizes Novel Candidate Genes from Recent Literature To assess the N-CS’s performance on real-world data, we tested it on a validation set of 46 variants from 11 novel candidate genes identified in recent literature (Supplementary Methods, Section S5, Supplementary Results, Section S9). This set comprised candidate genes for a range of inherited kidney diseases, including five for CAKUT ( ARID3A [ 20 ], NR6A1 [ 20 , 21 ], SOX13 [ 22 ], PHIP [ 23 ], and TSHZ3 [ 24 ], PIEZO [ 25 ]), two for PKD ( UGGT1 [ 26 ] and CFAP47 [ 27 ]) and individual candidates for ciliopahthies ( WDR44 [ 28 ]), a candidate gene for kidney displasia mimicking salt-wasting ( TFCP2L1 [ 29 ]), and proteinuric kidney disease ( MYO1C [ 30 ]). The N-CS assigned high scores to the majority of these newly discovered candidates, demonstrating its ability to rank them appropriately ( Fig. 4 ). Download figure Open in new tab Figure 4. N-CS on novel candidate genes from a systematic literature review. The plot displays N-CS scores for 46 variants across 11 recently published candidate genes. Each point represents a single variant, colored by its mode of inheritance. Variants in 10 of the 11 novel genes achieved scores > 7. These scores were observed across various inheritance patterns, including de novo , dominant, and X-linked recessive modes. For example, novel candidates such as ARID3A, WDR44 , and SOX13 consistently received top-tier scores for their associated variants. In contrast, variants in CFAP47 received comparatively lower scores. Discussion Current nephrogenetics faces a significant challenge in identifying causal genetic variants among a large amount of harmless background noise. This problem is particularly pertinent in adult kidney disease, which, in contrast to many pediatric disorders, is often characterized by late onset, incomplete penetrance, and variable expressivity that obscure clear genotype-phenotype correlations and have historically hampered gene discovery. To overcome this long-standing challenge, we created and validated the N-CS, a comprehensive ML-driven framework for automated candidate variant prioritization. Our findings show that by combining gene-specific biological relevance (N-GS), variant-level deleteriousness (N-VS), and mode of inheritance (IS), the N-CS provides a reliable and standardized method for sifting through large-scale sequencing data. While other computational tools for kidney genetics exist, the N-CS was developed to fill a specific niche. Its framework was informed by a similar tool - AutoCaSc - that was previously developed by some of us for neurodevelopmental disorders [ 31 ]. However, genetic kidney diseases present distinct challenges compared to severe pediatric phenotypes. Kidney disorders frequently manifest in adulthood, often after reproductive age, meaning causal variants are under weaker negative selection and are thus harder to distinguish from benign variation. These complexities, along with reduced penetrance and variable expressivity, suggest that traditional, rule-based scoring systems may fail to capture the subtle patterns of pathogenicity. We therefore specifically designed the N-CS to use machine learning (ML), which can detect intricate patterns within complex biological data that might be missed by human-defined heuristics. This approach distinguishes the N-CS from other valuable resources in kidney genetics. For instance, KidneyGPS is a web-based application designed to help researchers navigate the results of large-scale genome-wide association studies (GWAS) for kidney function. It provides an interactive platform to explore genes and variants within 424 eGFR-associated loci (594 independent signals), integrating fine-mapping, QTL data, phenotype associations, and druggability information to prioritize candidates involved in kidney function and CKD risk [ 32 ]. In contrast, the N-CS is designed to analyze rare variants from exome or genome sequencing. A key advantage of this approach is that it is not restricted to known loci; it performs an unbiased, exome-wide analysis to prioritize variants in any gene, offering a much broader scope for novel gene discovery. Another tool, KidneyNetwork , prioritizes novel candidate genes by leveraging a kidney-specific co-expression network, operating on the principle of “guilt by association” (the idea that genes with similar expression patterns likely share related biological functions) [ 33 ]. While also focused on rare disease gene discovery, its methodology is primarily based on an unsupervised analysis of RNA-sequencing patterns. The N-CS differs significantly by employing a supervised ML model trained on a broad spectrum of 109 genomic features, including conservation, population constraint metrics, and mouse model data, in addition to gene expression. The key advantage of this ML approach is its ability to learn the complex, non-linear relationships and subtle weightings between these diverse features—patterns that are often too intricate to be captured by human-defined rules or methods based on a single data type, such as co-expression. Furthermore, the N-CS provides a more granular, multi-faceted final score that evaluates not only the gene (N-GS) but also the specific variant’s predicted deleteriousness (N-VS) and its mode of inheritance (IS), offering a comprehensive framework tailored to the clinical challenge of resolving individual genetic cases. A key strength of the N-CS is its potential to move beyond the highly penetrant monogenic models that are often insufficient to explain adult-onset kidney disease. By using the N-GS as a biological prior, the N-CS can substantially reduce the genetic search space, empowering association studies to detect signals of moderate effect even in smaller cohorts. This utility is exemplified by the distinct U-shaped distribution of the N-GS among novel genes, which effectively partitions thousands of uncharacterized genes into a large group of likely irrelevant candidates and a smaller, manageable subset of high-priority genes warranting further investigation. The utility of the N-CS was demonstrated by its strong performance in our literature-based validation experiment. By assigning high scores to the clear majority of recently published variants in novel CKD candidate genes, the framework proved its ability to generalize beyond its training data and identify clinically relevant signals as new discoveries are made. Primarily, the N-CS was designed for researchers as a hypothesis-generation engine, capable of pinpointing high-priority genes for functional follow-up and enabling more powerful cohort analyses. However, it could also serve as a crucial decision-support tool for clinical geneticists. By automatically prioritizing the hundreds of rare variants of uncertain significance (VUS) often identified by exome sequencing, it could allow experts to focus their deep-dive manual ACMG classification on a manageable handful of top candidates, potentially accelerating the diagnostic process and helping to resolve difficult cases. As a standardized and computable toolset, the N-CS makes the systematic re-analysis of the many existing sequencing cohorts feasible, allowing researchers to apply a consistent and comparable scoring logic to previously generated data. To ensure broad accessibility, we have released the N-CS as an open-source suite of tools designed to facilitate the standardized and automated assessment of candidate variants in genetic kidney diseases. We created a user-friendly public web tool ( https://nc-scorer.kidney-genetics.org/ ) and a command-line interface, encouraging community-wide application, sharing, and validation. This approach helps level the playing field, allowing smaller research groups without extensive bioinformatics support to conduct sophisticated gene discovery analyses and share their results in a comparable way. Ultimately, creating this open and standardized ecosystem benefits the entire research community and, most importantly, the patients waiting for a diagnosis. While the current N-CS provides a robust framework for prioritizing coding variants, its underlying methodology could be adapted to address further challenges in nephrogenetics. The framework could be extended to incorporate non-coding regulatory variants and structural variations, which are currently not assessed. Such an extension would require the integration of different genomic feature sets, such as data from ChIP-seq or chromatin conformation assays, to accurately model the functional impact of variants outside of protein-coding regions. Furthermore, the N-CS framework could be adapted to create phenotype- or disease subgroup-specific scores. For instance, separate models could be trained specifically for cystic kidney diseases, CAKUT, or glomerulopathies. Such specialized scores would likely offer enhanced predictive power and greater clinical specificity compared to the current pan-kidney score. There is also potential for integrating the N-CS with other emerging AI technologies. A promising application is the use of Large Language Models (LLMs) to conduct detailed evaluations of top-ranking candidates. After using the N-CS for high-throughput pre-filtering to generate a manageable shortlist, an LLM could perform an automated, in-depth literature review to synthesize existing evidence and assess the biological plausibility of a gene-disease link for each pre-selected candidate. Finally, the N-CS could be applied beyond the search for primary causal mutations to identify modifier genes in known monogenic conditions. By analyzing genomic data from large cohorts of patients with diseases like ADPKD, the N-CS could help pinpoint secondary genetic variants that influence disease severity, progression, or extra-renal manifestations, thus helping to explain the significant clinical variability observed in these disorders. Limitations This study has several limitations. The N-CS models were trained on currently known kidney disease genes and variants, which may bias the score against identifying genes that operate through entirely novel biological mechanisms. The tool’s performance is also inherently dependent on the quality and completeness of the external databases from which its features are derived. Additionally, the current N-CS framework is designed to evaluate single nucleotide variants and small indels, and is not equipped to assess the pathogenicity of non-coding regulatory variants or larger structural variations, which are also known causes of IKD. Conclusion The N-CS framework represents a significant step forward in navigating the complex genetic landscape of CKD. By integrating gene-level evidence, variant-level pathogenicity, and inheritance patterns into a single, robust machine learning-driven score, it provides a powerful tool helping overcome the challenges of late onset and incomplete penetrance that have historically hindered gene discovery. As a publicly available and user-friendly tool, the N-CS is designed to empower researchers and clinicians, accelerating the pace of discovery and shortening the diagnostic odyssey for patients. A crucial next step will be the large-scale validation of the N-CS in prospective clinical cohorts, such as the German Chronic Kidney Disease (GCKD) study [ 34 ], to conclusively determine its real-world clinical utility and generalizability. Data Availability Most data produced are available online at https://github.com/halbritter-lab/nephro_candidate_score. Due to large file sizes, some data is not openly accessible, but is available upon reasonable request to the authors. https://github.com/halbritter-lab/nephro_candidate_score Software and Data Availability The Nephro Gene Score (N-GS) features were acquired and preprocessed using R version 4.3.2, while the Nephro Variant Score (N-VS) features, as well as all subsequent machine learning model training and testing, were done in Python version 3.7. The Supplementary Material contains additional information about the specific data sources used to generate features. The complete source code, as well as most raw data files and all processed data files needed to replicate the analyses presented in this study, are publicly available on GitHub ( https://github.com/halbritter-lab/nephro_candidate_score ). The repository excludes some raw data files due to their large size. They can be downloaded directly from their respective public platforms (as described in the Supplementary Material) or obtained from the corresponding author on reasonable request. References 1. ↵ Foreman KJ , Marquez N , Dolgert A , Fukutaki K , Fullman N , McGaughey M , et al. Forecasting life expectancy, years of life lost, and all-cause and cause-specific mortality for 250 causes of death: reference and alternative scenarios for 2016--40 for 195 countries and territories . Lancet . 2018 ; 392 : 2052 – 2090 . OpenUrl CrossRef PubMed 2. ↵ Titze S , Schmid M , Köttgen A , Busch M , Floege J , Wanner C , et al. Disease burden and risk profile in referred patients with moderate chronic kidney disease: composition of the German Chronic Kidney Disease (GCKD) cohort . Nephrol Dial Transplant . 2015 ; 30 : 441 – 451 . OpenUrl CrossRef PubMed 3. ↵ Zhang J , Thio CHL , Gansevoort RT , Snieder H. Familial aggregation of CKD and heritability of kidney biomarkers in the general population: The Lifelines Cohort Study . Am J Kidney Dis . 2021 ; 77 : 869 – 878 . OpenUrl PubMed 4. ↵ KDIGO Conference Participants . Genetics in chronic kidney disease: conclusions from a Kidney Disease: Improving Global Outcomes (KDIGO) Controversies Conference . Kidney Int . 2022 ; 101 : 1126 – 1141 . OpenUrl PubMed 5. ↵ Groopman E , Goldstein D , Gharavi A. Diagnostic utility of exome sequencing for kidney disease . Reply. N Engl J Med . 2019 ; 380 : 2080 – 2081 . OpenUrl PubMed 6. ↵ kidney-genetics-v1: Kidney-Genetics - database of kidney-related genes . Github ; Available: https://github.com/halbritter-lab/kidney-genetics-v1 7. ↵ Schönauer R , Baatz S , Nemitz-Kliemchen M , Frank V , Petzold F , Sewerin S , et al. Matching clinical and genetic diagnoses in autosomal dominant polycystic kidney disease reveals novel phenocopies and potential candidate genes . Genet Med . 2020 ; 22 : 1374 – 1383 . OpenUrl PubMed 8. ↵ van der Ven AT , Vivante A , Hildebrandt F. Novel Insights into the Pathogenesis of Monogenic Congenital Anomalies of the Kidney and Urinary Tract . J Am Soc Nephrol . 2018 ; 29 : 36 – 50 . OpenUrl Abstract / FREE Full Text 9. ↵ Halbritter J , Figueres L , Van Eerde AM , Capasso G , Hoorn EJ , Nijenhuis T , et al. Chronic Kidney Disease of unexplained cause (CKDx): a consensus statement by the Genes & Kidney Working Group of the ERA . Nephrol Dial Transplant . 2025 . doi: 10.1093/ndt/gfaf092 OpenUrl CrossRef 10. ↵ Richards S , Aziz N , Bale S , Bick D , Das S , Gastier-Foster J , et al. Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology . Genet Med . 2015 ; 17 : 405 – 424 . OpenUrl CrossRef PubMed 11. ↵ Karczewski KJ , Francioli LC , Tiao G , Cummings BB , Alföldi J , Wang Q , et al. The mutational constraint spectrum quantified from variation in 141,456 humans . Nature . 2020 ; 581 : 434 – 443 . OpenUrl CrossRef PubMed 12. ↵ Leitão E , Schröder C , Parenti I , Dalle C , Rastetter A , Kühnel T , et al. Systematic analysis and prediction of genes associated with monogenic disorders on human chromosome X . Nat Commun . 2022 ; 13 : 6570 . OpenUrl CrossRef PubMed 13. ↵ Amberger JS , Bocchini CA , Schiettecatte F , Scott AF , Hamosh A. OMIM.org: Online Mendelian Inheritance in Man (OMIM®), an online catalog of human genes and genetic disorders . Nucleic Acids Res . 2015 ; 43 : D789 – 98 . OpenUrl CrossRef PubMed 14. ↵ Pedregosa F , Varoquaux G , Gramfort A , Michel V , Thirion B , Grisel O , et al. Scikit-learn: Machine Learning in Python . J Mach Learn Res . 2011 ; 12 : 2825 – 2830 . OpenUrl CrossRef PubMed 15. ↵ Subramanian A , Tamayo P , Mootha VK , Mukherjee S , Ebert BL , Gillette MA , et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles . Proc Natl Acad Sci U S A . 2005 ; 102 : 15545 – 15550 . OpenUrl Abstract / FREE Full Text 16. ↵ Fang Z , Liu X , Peltz G. GSEApy: a comprehensive package for performing gene set enrichment analysis in Python . Bioinformatics . 2023 ; 39 . doi: 10.1093/bioinformatics/btac757 OpenUrl CrossRef PubMed 17. ↵ Landrum MJ , Lee JM , Riley GR , Jang W , Rubinstein WS , Church DM , et al. ClinVar: public archive of relationships among sequence variation and human phenotype . Nucleic Acids Res . 2014 ; 42 : D980 – 5 . OpenUrl CrossRef PubMed Web of Science 18. ↵ McLaren W , Gil L , Hunt SE , Riat HS , Ritchie GRS , Thormann A , et al. The Ensembl Variant Effect Predictor . Genome Biol . 2016 ; 17 : 122 . OpenUrl CrossRef PubMed 19. ↵ Wei C-H , Allot A , Lai P-T , Leaman R , Tian S , Luo L , et al. PubTator 3.0: an AI-powered literature resource for unlocking biomedical knowledge . Nucleic Acids Res . 2024 ; 52 : W540 – W546 . OpenUrl CrossRef PubMed 20. ↵ Milo Rasouly H , Krishna Murthy SB , Vena N , Povysil G , Beenken A , Verbitsky M , et al. Exome analysis links kidney malformations to developmental disorders and reveals causal genes . Nat Commun . 2025 ; 16 : 7290 . OpenUrl PubMed 21. ↵ Neelathi UM , Ullah E , George A , Maftei MI , Boobalan E , Sanchez-Mendoza D , et al. Variants in NR6A1 cause a novel oculo vertebral renal syndrome . Nat Commun . 2025 ; 16 : 6111 . OpenUrl PubMed 22. ↵ Merz LM , Kolvenbach CM , Wang C , Mertens ND , Seltzsam S , Mansour B , et al. Trio exome sequencing identifies de novo variants in novel candidate genes in 19.62% of CAKUT families . Genet Med . 2025 ; 27 : 101432 . OpenUrl PubMed 23. ↵ Rivera-Munoz EA , Zhao XE , Rosenfeld JA , Luna PN , Shaw CA , Posey JE , et al. Clinical exome sequencing efficacy and phenotypic expansions involving non-isolated congenital anomalies of kidney and urinary tract (CAKUT+) . Eur J Hum Genet . 2025 . doi: 10.1038/s41431-025-01929-3 OpenUrl CrossRef 24. ↵ Kesdiren E , Martens H , Brand F , Werfel L , Wedekind L , Trowe M-O , et al. Heterozygous variants in the teashirt zinc finger homeobox 3 (TSHZ3) gene in human congenital anomalies of the kidney and urinary tract . Eur J Hum Genet . 2025 ; 33 : 44 – 55 . OpenUrl PubMed 25. ↵ Amado NG , Nosyreva ED , Thompson D , Egeland TJ , Ogujiofor OW , Yang M , et al. PIEZO1 loss-of-function compound heterozygous mutations in the rare congenital human disorder Prune Belly Syndrome . Nat Commun . 2024 ; 15 : 339 . OpenUrl PubMed 26. ↵ Dardas Z , Harrold L , Calame DG , Salter CG , Kikuma T , Guay KP , et al. Bi-allelic UGGT1 variants cause a congenital disorder of glycosylation . Am J Hum Genet . 2025 ; 112 : 1139 – 1157 . OpenUrl PubMed 27. ↵ Mori T , Fujimaru T , Liu C , Patterson K , Yamamoto K , Suzuki T , et al. CFAP47 is implicated in X-linked polycystic kidney disease . Kidney Int Rep . 2024 ; 9 : 3580 – 3591 . OpenUrl PubMed 28. ↵ Accogli A , Shakya S , Yang T , Insinna C , Kim SY , Bell D , et al. Variants in the WDR44 WD40-repeat domain cause a spectrum of ciliopathy by impairing ciliogenesis initiation . Nat Commun . 2024 ; 15 : 365 . OpenUrl PubMed 29. ↵ Vaqueiro Graña M , Madariaga L , Gómez-Conde S , Iceta Lizarraga A , Hualde Olascoaga J , Ariceta G. Ultra-rare severe kidney dysplasia mimicking salt-wasting tubulopathy associated with TFCP2L1 gene variants . Pediatr Nephrol . 2025 . doi: 10.1007/s00467-025-06804-3 OpenUrl CrossRef 30. ↵ Elmubarak I , Shril S , Mansour B , Bao A , Kolvenbach CM , Kari JA , et al. Recessive variants in MYO1C as a potential novel cause of proteinuric kidney disease . Pediatr Nephrol . 2024 ; 39 : 2939 – 2945 . OpenUrl PubMed 31. ↵ Lieberwirth JK , Büttner B , Klöckner C , Platzer K , Popp B , Abou Jamra R. AutoCaSc: Prioritizing candidate genes for neurodevelopmental disorders . Hum Mutat . 2022 ; 43 : 1795 – 1807 . OpenUrl PubMed 32. ↵ Stanzick KJ , Stark KJ , Gorski M , Schödel J , Krüger R , Kronenberg F , et al. KidneyGPS: a user-friendly web application to help prioritize kidney function genes and variants based on evidence from genome-wide association studies . BMC Bioinformatics . 2023 ; 24 : 355 . OpenUrl PubMed 33. ↵ Boulogne F , Claus L , Wiersma H , Oelen R , Schukking F , de Klein N , et al. KidneyNetwork: Using kidney-derived gene expression data to predict and prioritize novel genes involved in kidney disease . doi: 10.21203/rs.3.rs-1870632/v1 OpenUrl CrossRef 34. ↵ Eckardt K-U , Bärthlein B , Baid-Agrawal S , Beck A , Busch M , Eitner F , et al. The German Chronic Kidney Disease (GCKD) study: design and methods . Nephrol Dial Transplant . 2012 ; 27 : 1454 – 1460 . OpenUrl CrossRef PubMed Web of Science View the discussion thread. Back to top Previous Next Posted September 30, 2025. Download PDF Supplementary Material Data/Code Email Thank you for your interest in spreading the word about medRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following Automatic variant prioritization in suspected genetic kidney disease using the Nephro Candidate Score (N-CS) Message Subject (Your Name) has forwarded a page to you from medRxiv Message Body (Your Name) thought you would like to see this page from the medRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share Automatic variant prioritization in suspected genetic kidney disease using the Nephro Candidate Score (N-CS) Nina Rank , Sören Lukassen , Manuel Anderegg , Kai-Uwe Eckardt , Jan Halbritter , Bernt Popp medRxiv 2025.09.29.25336840; doi: https://doi.org/10.1101/2025.09.29.25336840 Share This Article: Copy Citation Tools Automatic variant prioritization in suspected genetic kidney disease using the Nephro Candidate Score (N-CS) Nina Rank , Sören Lukassen , Manuel Anderegg , Kai-Uwe Eckardt , Jan Halbritter , Bernt Popp medRxiv 2025.09.29.25336840; doi: https://doi.org/10.1101/2025.09.29.25336840 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Nephrology Subject Areas All Articles Addiction Medicine (568) Allergy and Immunology (863) Anesthesia (300) Cardiovascular Medicine (4435) Dentistry and Oral Medicine (444) Dermatology (382) Emergency Medicine (608) Endocrinology (including Diabetes Mellitus and Metabolic Disease) (1509) Epidemiology (15228) Forensic Medicine (30) Gastroenterology (1124) Genetic and Genomic Medicine (6598) Geriatric Medicine (668) Health Economics (997) Health Informatics (4536) Health Policy (1368) Health Systems and Quality Improvement (1613) Hematology (540) HIV/AIDS (1264) Infectious Diseases (except HIV/AIDS) (15916) Intensive Care and Critical Care Medicine (1103) Medical Education (623) Medical Ethics (146) Nephrology (667) Neurology (6599) Nursing (346) Nutrition (998) Obstetrics and Gynecology (1144) Occupational and Environmental Health (957) Oncology (3332) Ophthalmology (974) Orthopedics (369) Otolaryngology (420) Pain Medicine (436) Palliative Medicine (130) Pathology (663) Pediatrics (1693) Pharmacology and Therapeutics (691) Primary Care Research (711) Psychiatry and Clinical Psychology (5447) Public and Global Health (9230) Radiology and Imaging (2198) Rehabilitation Medicine and Physical Therapy (1370) Respiratory Medicine (1196) Rheumatology (593) Sexual and Reproductive Health (712) Sports Medicine (530) Surgery (712) Toxicology (99) Transplantation (289) Urology (265) (function(){function c(){var b=a.contentDocument||a.contentWindow.document;if(b){var d=b.createElement('script');d.innerHTML="window.__CF$cv$params={r:'a005203fbe0df6ec',t:'MTc3OTU1MDQ3OQ=='};var a=document.createElement('script');a.src='/cdn-cgi/challenge-platform/scripts/jsd/main.js';document.getElementsByTagName('head')[0].appendChild(a);";b.getElementsByTagName('head')[0].appendChild(d)}}if(document.body){var a=document.createElement('iframe');a.height=1;a.width=1;a.style.position='absolute';a.style.top=0;a.style.left=0;a.style.border='none';a.style.visibility='hidden';document.body.appendChild(a);if('loading'!==document.readyState)c();else if(window.addEventListener)document.addEventListener('DOMContentLoaded',c);else{var e=document.onreadystatechange||function(){};document.onreadystatechange=function(b){e(b);'loading'!==document.readyState&&(document.onreadystatechange=e,c())}}}})();