{"paper_id":"8a57d6fb-8fb4-459a-b37d-60089b677454","body_text":"Uterine microbiome signatures associated with endometriosis | 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 Uterine microbiome signatures associated with endometriosis Libo Zhu, Jiaying He, Xiaochun Xu, Shen Lu, Yanqin Yu, Wing Hing Wong, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7016822/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 10 You are reading this latest preprint version Abstract Background Endometriosis is a chronic inflammatory disorder affecting ~ 10% of reproductive-age women, often causing pelvic pain and infertility. Despite its prevalence, diagnosis remains delayed due to non-specific symptoms and lack of reliable non-invasive biomarkers. Emerging evidence implicates the microbiome in disease pathogenesis. Results We analyzed uterine microbiomes from 266 tissue samples collected during either the proliferative or secretory phase, using 16S rRNA gene sequencing. Genus-level analysis revealed variable Lactobacillus abundance among all individuals. Prevotella showed borderline enrichment in proliferative-phase patients. Sub-genus analyses identified a small number of differentially abundant taxa, though none remained significant after FDR correction. To capture subtle microbial shifts, we developed a feature set combining weakly differential taxa, algorithmically selected taxa via machine-learning, and a functional dysbiosis score. A supervised classifier trained on proliferative-phase data achieved moderate predictive performance (AUC = 0.70), while secretory-phase models performed poorly (AUC = 0.58). Conclusion The uterine microbiome shows phase-dependent differences in its potential to inform endometriosis status. Although no robust individual microbial biomarkers were identified, machine-learning models incorporating subtle community features from the proliferative phase yielded modest diagnostic potential. These results highlight the importance of menstrual cycle-aware sampling and support further development of microbiome-informed diagnostic tools for endometriosis. Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Background Endometriosis is a chronic inflammatory condition where endometrial-like tissue grows outside the uterus, often affecting the ovaries, fallopian tubes, and peritoneum ( 1 , 2 ). This ectopic tissue responds to hormonal changes, leading to symptoms like chronic pelvic pain, dysmenorrhea, dyspareunia, and infertility ( 1 – 2 ). Affecting an estimated 10% of reproductive-age women, the disease is frequently underdiagnosed, with an average delay of 6–11 years ( 3 ). This is due to its non-specific symptoms, overlap with other conditions, and the lack of reliable diagnostics. Multiple theories exist about the origin of endometriosis – including retrograde menstruation, coelomic metaplasia, and stem cell involvement ( 4 ). Immune and inflammatory responses are also thought to play a key role, and recent research has begun exploring the gut and vaginal microbiomes as potential contributors ( 5 ). Studies show that women with endometriosis often have higher levels of Escherichia coli in the gut, linked to elevated serum lipopolysaccharide, which activates inflammatory pathways and promotes pro-inflammatory cytokine production. These cytokines may aid the survival and implantation of ectopic endometrial cells ( 6 ). Additionally. Streptococcus species have been linked to advanced disease stages, possibly by stimulating prostaglandin E2 expression, a contributor to pelvic pain ( 7 ). However, findings across studies are inconsistent; notably, one large study involving around 1000 participants found no significant association between gut microbiome and endometriosis ( 8 ). The vaginal microbiome, crucial for reproductive health, remains relatively understudied in the context of endometriosis ( 9 ). Shifts in vaginal microbiome, particularly the reduction of Lactobacillus and overgrowth of bacteria such as Gardnerella , Prevotella and Mobiluncus are seen in bacterial vaginosis (BV), which is associated with inflammation and may contribute to endometriosis pathogenesis ( 10 ). Moreover, menstrual cycle phases influence vaginal microbial composition, with reduced Lactobacillus and increased diversity during menstruation, complicating research and emphasizing the importance of cycle-aware sampling ( 11 ). To investigate this further, we analyzed uterine microbiomes in 266 samples from women in either the proliferative or secretory phases. A total of 138 uterine tissue samples were collected from women in the proliferative phase of their menstrual cycle. Among these, 78 samples were obtained from women with a laparoscopic diagnosis of endometriosis, while the remaining 60 were from women without the disease. An additional 128 uterine tissue samples were collected during the secretory phase, including 88 from women diagnosed with endometriosis and 40 from unaffected individuals (Table 1 ; Fig. 1 ). Total genomic DNA was extracted from all tissue samples and used to prepare targeted bacterial 16S rRNA gene libraries for sequencing, as detailed in the Methods section. Raw sequence data were processed to remove technical artifacts, host DNA, and environmental contaminants. High-confidence bacterial reads were then taxonomically annotated using an internally curated version of the Greengenes2 database. Downstream analyses focused on identifying differentially abundant microbial taxa associated with endometriosis and determining taxa with potential predictive value for disease diagnosis. Table 1 Summary of study samples. Median values are presented for each demographic parameter, with ranges shown in brackets. Statistical comparisons were conducted using T-tests. 1. Characteristics Endometriosis Control p -value All samples (proliferative and secretory phases) Sample size 166 100 - BMI 21.48 (15.62–36.85) 22.50 (17.1–31.22) 0.83 Age 35.5 (20–51) 38.5 (21–50) 0.0002 Proliferative phase samples Sample size 78 60 - BMI 21.51 (16.21–36.85) 22.50 (17.1–30.42) 0.76 Age 36 (20–51) 40.5 (24–50) 0.026 Secretory phase samples Sample size 88 40 - BMI 21.45 (15.62–34.22) 22.49 (18.22–31.22) 0.94 Age 35 (21–51) 37.5 (21–49) 0.005 Results Uterine microbiome landscape in the study cohort The initial objective of our analysis was to evaluate both alpha and beta diversity of the uterine microbiome in women diagnosed with endometriosis compared to those without the disease, stratified by the proliferative and secretory phases of the menstrual cycle. Alpha diversity, which reflects the richness and evenness of microbial species within individual samples, was assessed using the Shannon Index. No statistically significant differences in alpha diversity were observed between endometriosis and control groups in either the proliferative or secretory phase (Fig. 2 a). Similarly, beta diversity, which measures compositional differences in microbial communities between groups, was evaluated using the Bray-Curtis dissimilarity metric. This analysis also revealed no significant differences between women with and without endometriosis across both menstrual phases (Fig. 2 b). These findings suggest that the overall diversity, including both the number of microbial taxa and their relative abundance distribution, is comparable between affected and unaffected individuals, irrespective of the menstrual cycle phase. Genus-level analysis revealed substantial variability in the relative abundance of Lactobacillus among individuals, both in patients and controls, across both menstrual phases. Although Lactobacillus is typically considered a hallmark of a healthy vaginal microbiome, its levels varied considerably, especially among patients with endometriosis in the proliferative phase compared to the controls (Fig. 3 a-b), recapitulating a previous study which showed considerable variations in Lactobacillus abundance even among healthy females ( 11 ). An analysis of the top 20 most abundant genera in both proliferative and secretory phase samples showed that the overall distribution of relative abundance was similar between patients and controls, with most of these dominant taxa not differentially abundant. One notable observation was made: the genus Prevotella showed a trend toward enrichment in patient samples from the proliferative phase, with a borderline significant p-value (0.0509). Bacteria species in the genus Prevotella are commonly associated with vaginal dysbiosis and pro-inflammatory states ( 10 ). Differentially abundant and machine-learning informative taxa A more in-depth analysis of sub-genus level taxonomic units revealed eight taxa that were differentially abundant (p-value ≤ 0.05) between patients and controls in the proliferative phase, and three differential taxa in the secretory phase (Fig. 4 a), after adjusting for potential confounding effects of BMI and age. There is no overlap in differential taxa between proliferative and secretory phases. Although the number of differentially abundant taxa was modest, this finding aligns with expectations based on a prior large-scale study where it reported no statistically significant differences in gut microbial composition between women with and without endometriosis ( 8 ), despite earlier, smaller studies suggesting such associations. This discrepancy highlights the challenges in identifying consistent microbial biomarkers of gynecological disease, particularly in extraintestinal sites. Motivated by the hypothesis that the uterine microbiome may more directly reflect gynecological pathophysiology than the gut microbiome, we undertook this study to examine microbial community profiles in uterine tissue. In fact, after correcting for multiple comparisons using false discovery rate (FDR) adjustment, the initially observed differential taxa no longer reached statistical significance (i.e. FDR > 0.05). Nonetheless, we recognize that subtle yet consistent shifts in microbial composition across multiple taxa may carry predictive value ( 12 ). Therefore, we aim to employ machine learning approaches to integrate these signals, under the premise that the cumulative effect of multiple weakly informative taxa would enable and/or enhance predictive performance in distinguishing disease states ( 13 ). For instance, taxa such as Prevotella sp.1 and Ureaplasma sp.1 – both belonging to genera frequently linked to bacterial vaginosis ( 10 ), while not individually conclusive after FDR correction, may collectively contribute to distinguishing disease states when incorporated as features in a supervised machine learning model. To develop the feature set for supervised machine learning classification, we implemented a three-step selection strategy combining statistical and algorithmic criteria. First, we identified weakly differential taxa (i.e. nominal p-values ≤ 0.05; Fig. 4 a & 4 b), indicating potential biological relevance despite not meeting strict multiple-testing thresholds. These taxa were initially included to ensure that subtle, non-random differences were not overlooked. In the second step, we applied a machine learning-based feature selection process by systematically evaluating the importance of each taxon detected in our profiling pipeline. This involved training preliminary models to score each taxon’s contribution to classification performance, using a predefined threshold of feature importance score ≥ 0.015 as a cutoff for inclusion. Taxa meeting this threshold were selected as additional candidates for the final feature set (Fig. 4 c; Supplementary Table S1 -S2). A subset of taxa was identified exclusively through the feature importance criterion. Specifically, 14 additional taxa including two other Prevotella spp. were added in the proliferative phase. In the secretory phase, 11 additional taxa were incorporated. Notably, Gardnerella sp.1, a taxon traditionally linked to bacterial vaginosis was included in both proliferative and secretory cohorts. In the third step, a functional dysbiosis score (FDS) was calculated for each sample (Supplementary Table S3), representing an aggregate measure of microbial dysbiosis within the uterine tissue. The final feature set used for model training therefore comprised three components: the weakly differential taxa, taxa selected by feature importance scoring, and the FDS (Supplementary Table S4-S5). Diagnostic performance of machine-learning models in endometriosis detection Using the microbial profiles from the proliferative phase, we achieved moderate predictive performance in distinguishing endometriosis patients from controls. Specifically, across 50 rounds of repeated random subsampling cross-validation, the average AUC reached 0.70, indicating reasonable discriminative capability. The model demonstrated a sensitivity of 0.71, and a specificity of 0.54 (Fig. 2 d). While not optimal, the overall performance suggests that the microbiome during the proliferative phase carries meaningful signals that could aid in endometriosis diagnosis, especially when combined with other tools such as laparoscopy and blood-based biomarker tests. In contrast, models trained on the microbial profiles from secretory phase showed weaker overall performance, with an average AUC of only 0.58 (Fig. 5 ). Overall, these findings suggest that microbial signatures differ in both menstrual phases, and that the proliferative phase carries more informative profiles. From a clinical and biological perspective, these results underscore the importance of menstrual cycle timing when considering the microbiome as a diagnostic aid for endometriosis. The result also suggests that cycle-phase-specific sampling may be crucial for optimizing microbiome-based diagnostics, and that future models may benefit from integrating hormonal phase information, or adjusting for it explicitly. Discussion This study investigated the uterine microbiome in women with and without endometriosis, with a focus on menstrual cycle phase-specific microbial signatures and their predictive value for disease status. The differentially abundant taxa, coupled with the use of supervised machine-learning allowed us to uncover subtle patterns of microbial variation that may hold diagnostic potential. Our findings add to a growing body of literature suggesting that microbial dysbiosis may contribute to endometriosis pathogenesis, though not through overt community-level disruption. Previous studies have implicated the gut and vaginal microbiomes, such as the elevated abundance of Gardnerella and Streptococcus being associated with advanced disease. However, large-scale studies such as the one by Perez-Prieto et al (2024), have failed to find significant associations between gut microbiome and endometriosis despite other studies suggesting so, indicating that location-specific microbial assessments may be more informative. Our results demonstrate that the uterine microbiome may contain weak but biologically meaningful signals, especially during the proliferative phase, where the tissue is more hormonally responsive and immunologically active. The modest predictive power (AUC = 0.7) achieved by our model trained on proliferative phase cohort is consistent with recent studies showing that cumulative patterns of multiple taxa can outperform single-biomarker approaches. Furthermore, the menstrual cycle phase significantly influences reproductive tract microbiota, with menstruation being associated with reduced Lactobacillus and increased microbial diversity. These fluctuations likely contribute to inconsistencies across studies in the literature and emphasize the importance of phase-aware sampling. Our observation that microbial profiles from the secretory phase were less predictive supports this notion and aligns with reports that progesterone-dominant conditions may dampen inflammatory signals or microbiome-host interactions. Conclusion In conclusion, while the uterine microbiome alone is unlikely to serve as a stand-alone biomarker for endometriosis, our findings suggest that when analyzed in cycle-phase-aware and integrative manner, it may contribute to a broader diagnostic framework. Future studies should explore longitudinal sampling, integrate host transcriptomic and immunologic data, assess signals in vaginal mucus, and validate predictive models in larger, independent cohorts. Collectively, these efforts may pave the way toward a microbiome-informed, minimally invasive diagnostic tools for endometriosis. Materials and Methods Specimen collection This study was approved by the institutional review board of the Women’s Hospital, Zhejiang University School of Medicine (IRB-20240110-R). Endometrial tissue samples were collected from 266 individuals, all of whom were clinically suspected of having a gynecologic condition and scheduled for laparoscopy with histopathological evaluation. To explore menstrual cycle-related differences, samples were obtained from women in either the proliferative or secretory phases of their cycle. The menstrual phase was initially assessed by physicians or surgeons based on self-reported cycle days and clinical evaluations. To confirm this classification, serum progesterone levels were measured using a protein assay from Kangrun Biotech Co. Ltd. (Guangdong, China), with levels above 1.08 ng/mL indicating the secretory phase, as per the manufacturer’s guidelines. Among the samples, 138 were from the proliferative phase (78 from individuals with endometriosis and 60 from controls), and 128 were from the secretory phase (88 with endometriosis and 40 controls). Endometriosis was diagnosed and confirmed via gold-standard laparoscopic surgery. All participants provided written informed consent. The collected tissue samples were immediately transported at 4°C to the Heranova Lifesciences laboratory and stored at -20°C upon arrival. Uterine tissue processing and targeted 16S library preparations A 3–5 mm fragment of endometrial tissue from each sample was placed into an individual centrifuge tube with 20 µL of Proteinase K and 180 µL of Buffer ATL. The mixture was vortexed thoroughly and incubated at 58°C with shaking at 1200 rpm for 3 hours. after incubation, add 210 µL of Buffer ATL to each sample and homogeniz them using the TissueLyser II (2 minutes at 30 Hz, 1-minute pause, repeated for 15 cycles), and DNA was extracted using the QIAsymphony SP instrument (QIAGEN, 35459). DNA concentration and purity were assessed using the MultiSkan GO spectrophotometer (Thermo, 1510). The V4 variable region of the 16S rRNA was amplified using Invitrogen Platinum SuperFi II DNA Polymerase with the following PCR conditions: initial denaturation at 98°C for 30 seconds (1 cycle), followed by 30 cycles of 98°C for 10 seconds, 60°C for 10 seconds, and 72°C for 30 seconds, and a final extension at 72°C for 5 minutes before holding at 4°C. The forward primer is 5’- TAATTGTGTGCCAGCMGCCGCGGTAA-3’ while the reverse primer is 5’- TCAGCCGGACTACHVGGGTWTCTAAT-3’. The PCR products were purified using VAHTS DNA Clean Beads. Adapter ligation was carried out using the UltraClean Universal DNA Library Prep Kit for Illumina V3 (Vazyme, UND607-02). First, 45 µL of End Repair reaction mix was added to the purified PCR product, followed by incubation at 20°C for 15 minutes, 65°C for 15 minutes, and held at 4°C. The ligation reaction mix was prepared on ice, added to the end-repaired DNA, and incubated at 20°C for 15 minutes, then held at 4°C. The ligated products were purified again with VAHTS DNA Clean Beads. Library amplification was performed under the following thermal conditions: 95°C for 3 minutes (1 cycle), then 5 cycles of 98°C for 20 seconds, 60°C for 15 seconds, and 72°C for 30 seconds, with a final extension at 72°C for 5 minutes and a hold at 4°C. The final libraries were purified using VAHTS DNA Clean Beads. Library concentrations were quantified using the KAPA Library Quantification Kit (KAPA, KK4824), and fragment sizes were evaluated with the Agilent 4200 TapeStation (Agilent, G2991A). Sequencing was performed on the Illumina MiSeq platform using the MiSeq Reagent Kit v2 (300 cycles). Bioinformatic processing of targeted 16S sequencing data The demultiplexed FASTQ files from Illumina MiSeq sequencing were processed to extract the forward reads. To improve data quality, a two-step trimming and filtering process was employed. First, fastp ( 14 ) was used to identify and remove polyX artifacts—artificial stretches of a single nucleotide—commonly introduced during sequencing. Next, cutadapt ( 15 ) was used to trim any residual adapter sequences from the reads. To eliminate host-derived contamination, the filtered reads were aligned to the human reference genome (hg38) using Bowtie2 ( 16 ). Alignment results were processed with SAMtools ( 17 ), and reads mapping to the human genome were removed, ensuring that only non-host (primarily bacterial) sequences were retained for microbiome analysis. The resulting high-quality, non-human reads were then imported into the QIIME2 platform ( 18 ) for microbial community analysis. Within QIIME2, chimeric sequences were identified and removed using the vsearch uchime-denovo method. Subsequently, redundant sequences were collapsed using vsearch dereplicate-sequences, enhancing computational efficiency and reducing noise. The remaining high-confidence bacterial reads were annotated using an internally curated version of the Greengenes2 reference database ( 19 ). Taxonomic assignments were made using the Greengenes2 taxonomy-from-table classifier, providing genus-level and species-level annotations where possible. To ensure the validity and accuracy of the microbiome profiles, decontamination was performed using SCRuB ( 20 ), a statistical tool designed to identify and remove background contaminants. A blank negative control, which underwent the entire experimental workflow alongside the tissue samples, was included in the analysis to model and subtract any environmental or reagent-based contaminants. This approach ensured that the final dataset reflected true biological signals and minimized the risk of false microbial detection. Microbiome Shannon index and beta diversity were calculated and visualized using vegan ( 21 ) in R (v.4.4.2). A functional dysbiosis score was computed for each sample using the following formula (0.5*(1-Lactobacillus) + 10*(Pathogenic taxa)) where pathogenic taxa consist of genus commonly associated bacterial vaginosis including Gardnerella , Prevotella , Anaerococcus , Streptococcus , Megasphaera , Mobiluncus , Sneathia , Atopobium , Peptoniphilus , Mycoplasmoides , Ureaplasma , Bacteroides , Peptostreptococcus and Dialister . Disease prediction model construction using a random forest classifier Samples were divided into proliferative (n = 138) and secretory (n = 128) groups for the following analysis, and the subsequent analysis was based on bacterial species relative abundances. MaAsLin2 ( 22 ) was performed to determine the multivariable association between bacterial species and endometriosis/non-endometriosis groups ( P ≤ 0.05), with age and BMI controlled using R package MaAsLin2. Features with importance scores ≥ 0.015 to the endometriosis/non-endometriosis groups among bacterial species were selected via random forest implemented in Python sklearn package. Models to predict endometriosis/non-endometriosis were built based on features that were selected by MaAsLin2, by random forest feature scoring and by the addition of the functional dysbiosis score. Model performance was assessed through 50 iterations of repeated random subsampling cross-validation, in which the data was randomly split into 80% for training and 20% for testing in each iteration. This strategy helped account for variability arising from random data splits and yielded a more robust estimate of the predictive accuracy. Declarations Ethics approval and consent to participate This study was approved by the institutional review board of the Women’s Hospital, Zhejiang University School of Medicine (IRB-20240110-R), and consent was given by enrolled participants. Consent for publication The enrolled participants provided informed consent for the use of their biological materials in research and publication. All authors have reviewed and approved this manuscript for publication. Data availability Sequencing data were deposited into Genome Sequence Archive (accession number PRJCA041172) Conflict-of-interest statement All authors, except Zhang Xinmei and Zhu Libo, are employees of Heranova Lifesciences, a company engaged in the commercial development of a non-invasive test for endometriosis. Farideh Bischoff holds stock options of Heranova Lifesciences. The authors declare no other conflicts of interest, financial or otherwise. Funding The research was performed using internal R&D budget at Heranova Lifesciences Inc. Author contributions L. Zhu and J. He analyzed the data. L. Zhu, J. He, W.H. Wong, F. Bischoff, X. Zhang conceptualized the study. L. Zhu, J. He and W.H. Wong wrote the manuscript with input from all authors. X. Zhang and L. Zhu provided the samples for sequencing. S. Lu coordinated sample collection and shipment to the laboratory. Y. Yu and X. Xu performed molecular experiments. W.H. Wong, F. Bischoff and X. Zhang supervised the study. F. Bischoff and X. Zhang secured R&D funding for the study. Acknowledgments The authors would like to express their appreciation to Jonathan Zhao and Frank Zhang (both from Heranova Lifesciences) for their insightful input on the study design and manuscript. References Zondervan KT, et al. Endometriosis. NEJM . 2020;382:1244-1256 Parasar P, et al. Endometriosis: Epidemiology, diagnosis and clinical management. Curr Obstet Gynecol Rep . 2017;6(1):34-41 Kirk UB, et al. Understanding endometriosis underfunding and its detrimental impact on awareness and research. npj Women’s Health . 2024;2:45 Lamceva J, et al. The main theories on the pathogenesis of endometriosis. International Journal of Molecular Sciences . 2023;24(5):4254 Ser H, et al. Current updates on the role of microbiome in endometriosis: a narrative review. Microorganisms . 2023;11(2):360 Tavana Z, et al. Significant increased isolation of Escherichia coli in Iranian women with endometriosis: a case-control study. BMC Women’s Health . 2024;24:383 Shan J, et al. Gut microbiota imbalance and its correlations with hormone and inflammatory factors in patients with stage 3/4 endometriosis. Arch Gynecol Obstet . 2021;304(5):1363-1373 Perez-Prieto I, et al. Gut microbiome in endometriosis: a cohort study on 1000 individuals. BMC Medicine . 2024;22:294 Brennan C, et al. Harnessing the power within: engineering the microbiome for enhanced gynecologic health. Reproduction and Fertility . 2024;5(2):e230060 Ding C, et al. Bacterial Vaginosis: effects on reproduction and its therapeutics. J Gynecol Obstet Hum Reprod . 2021;50(9):102174 Song SD, et al. Daily vaginal microbiota fluctuations associated with natural hormonal cycle, contraceptives, diet and exercise. mSphere . 2020;5(4):e00593-20 Chang D, et al. Gut microbiome Wellness Index 2 enhances health status prediction from gut microbiome taxonomic profiles. Nature Communications . 2024;15:7447 Wang X, et al. Metagenomics reveals unique gut mycobiome biomarkers in major depressive disorder – a non-invasive method. Frontiers in Cellular and Infection Microbiology . 2025;15:1582522 Chen S, et al. fastp: an ultra-fast all-in-one FASTQ preprocessor. Bioinformatics . 2018;34(17):i884-i890 Marcel M. Cutadapt removes adapter sequences from high-throughput sequencing reads. EMBnet.journal . 2011;17(1):10-12 Langmead Ben & Salzberg SL. Fast gapped-read alignment with Bowtie 2. Nature Methods . 2012;9:357-359 Li H, et al. The sequence alignment/map format and SAMtools. Bioinformatics . 2009;25(16):2078-2079 Bolyen E, et al. Reproducible, interactive, scalable and extensible microbiome data science using QIIME 2. Nature Biotechnology . 2019;37:852-857 McDonald D, et al. Greengenes2 unifies microbial data in a single reference tree. Nature Biotechnology . 2024;42:715-718 Austin GI, et al. Contamination source modeling with SCRuB improves cancer phenotype prediction from microbiome data. Nature Biotechnology . 2023;41:1820-1828 Dixon P. VEGAN, a package of R functions for community ecology. Journal of Vegetation Science . 2003;14(6):927-930 Mallick H, et al. Multivariable association discovery in population-scale meta-omics studies. PLoS Computational Biology . 2021;17(11):e100944 Additional Declarations Competing interest reported. All authors, except Zhang Xinmei and Zhu Libo, are employees of Heranova Lifesciences, a company engaged in the commercial development of a non-invasive test for endometriosis. Farideh Bischoff holds stock options of Heranova Lifesciences. The authors declare no other conflicts of interest, financial or otherwise. Supplementary Files SupplementaryMaterials.xlsx Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 19 Dec, 2025 Reviews received at journal 14 Dec, 2025 Reviewers agreed at journal 02 Dec, 2025 Reviews received at journal 29 Sep, 2025 Reviewers agreed at journal 29 Aug, 2025 Reviewers agreed at journal 23 Jul, 2025 Reviewers invited by journal 08 Jul, 2025 Editor assigned by journal 01 Jul, 2025 Submission checks completed at journal 01 Jul, 2025 First submitted to journal 01 Jul, 2025 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-7016822\",\"acceptedTermsAndConditions\":true,\"allowDirectSubmit\":false,\"archivedVersions\":[],\"articleType\":\"Research Article\",\"associatedPublications\":[],\"authors\":[{\"id\":482432731,\"identity\":\"a95d4a73-acff-4be3-bd82-e424ac6a5790\",\"order_by\":0,\"name\":\"Libo Zhu\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Zhejiang University\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Libo\",\"middleName\":\"\",\"lastName\":\"Zhu\",\"suffix\":\"\"},{\"id\":482432733,\"identity\":\"3af9cd7a-1026-4130-a57d-7e88d8b6aaa1\",\"order_by\":1,\"name\":\"Jiaying He\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Heranova Lifesciences\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Jiaying\",\"middleName\":\"\",\"lastName\":\"He\",\"suffix\":\"\"},{\"id\":482432735,\"identity\":\"4ad7cc88-1824-4932-86d0-593bec48489d\",\"order_by\":2,\"name\":\"Xiaochun Xu\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Heranova Lifesciences\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Xiaochun\",\"middleName\":\"\",\"lastName\":\"Xu\",\"suffix\":\"\"},{\"id\":482432737,\"identity\":\"290d75eb-aaea-4a05-9deb-b432c34945d9\",\"order_by\":3,\"name\":\"Shen Lu\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Heranova Lifesciences\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Shen\",\"middleName\":\"\",\"lastName\":\"Lu\",\"suffix\":\"\"},{\"id\":482432738,\"identity\":\"0366d24a-d7d3-4a23-ad45-9b85c5dcc61a\",\"order_by\":4,\"name\":\"Yanqin Yu\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Heranova Lifesciences\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Yanqin\",\"middleName\":\"\",\"lastName\":\"Yu\",\"suffix\":\"\"},{\"id\":482432739,\"identity\":\"6cac1f5f-0179-4fbd-82a3-abeb8839b9db\",\"order_by\":5,\"name\":\"Wing Hing Wong\",\"email\":\"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA9klEQVRIiWNgGAWjYDACZiBmbAAzGR+TrIXZmHiboFrYpIlSbXCc9/ALxh0Mif3S7deqCyruyZuz9xh+YKixicap5TBfmgXjGYbEmXPOlN2ecabYcGfPGWMJhmNpuQ04tJgd5jEzYGxjyN1wIyftNm9bAuOGG2kJEowNh4nTUsz7L8EeqCX5BwEtxg8gWtKPMfM2JCRuuJF8DK8t9kBbGBLbJOpnzshhluY5lpC84czhYxYJePwi2X/G+MPHNhtjfon0h595ahJsNxxvbL7xocYGpxYgYJNIYJAA0jwGCLEE3MpBgPkDhGZ/gF/dKBgFo2AUjFgAANKFWsqVopCgAAAAAElFTkSuQmCC\",\"orcid\":\"\",\"institution\":\"Heranova Lifesciences\",\"correspondingAuthor\":true,\"prefix\":\"\",\"firstName\":\"Wing\",\"middleName\":\"Hing\",\"lastName\":\"Wong\",\"suffix\":\"\"},{\"id\":482432740,\"identity\":\"eb2bfcbe-3dde-46e8-91fb-327f8030334f\",\"order_by\":6,\"name\":\"Farideh Z Bischoff\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Heranova Lifesciences\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Farideh\",\"middleName\":\"Z\",\"lastName\":\"Bischoff\",\"suffix\":\"\"},{\"id\":482432742,\"identity\":\"1506bbaa-522c-47df-9047-29a562f43ecb\",\"order_by\":7,\"name\":\"Xinmei Zhang\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Zhejiang University\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Xinmei\",\"middleName\":\"\",\"lastName\":\"Zhang\",\"suffix\":\"\"}],\"badges\":[],\"createdAt\":\"2025-07-01 06:53:09\",\"currentVersionCode\":1,\"declarations\":\"\",\"doi\":\"10.21203/rs.3.rs-7016822/v1\",\"doiUrl\":\"https://doi.org/10.21203/rs.3.rs-7016822/v1\",\"draftVersion\":[],\"editorialEvents\":[],\"editorialNote\":\"\",\"failedWorkflow\":false,\"files\":[{\"id\":86656106,\"identity\":\"dd9149a0-afd8-4ed6-a2f0-231fe47253e6\",\"added_by\":\"auto\",\"created_at\":\"2025-07-14 10:17:30\",\"extension\":\"png\",\"order_by\":1,\"title\":\"Figure 1\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":84554,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003eOverview of the study design. A total of 266 samples were analyzed, including 138 from participants in the proliferative phase (78 patients and 60 controls) and 128 from the secretory phase (88 patients and 40 controls). Targeted sequencing of the 16S rRNA V4 hypervariable region was performed. Bacterial reads were annotated using the Greengenes2 database. Differential and informative taxa were identified for machine learning-based prediction of endometriosis.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"1.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-7016822/v1/92180301344ead9c1e671836.png\"},{\"id\":86656105,\"identity\":\"d106d5b2-dae6-4322-ba53-218e5eb0b675\",\"added_by\":\"auto\",\"created_at\":\"2025-07-14 10:17:30\",\"extension\":\"png\",\"order_by\":2,\"title\":\"Figure 2\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":82373,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003e(a) Alpha diversity comparisons between patients and controls within each menstrual phase. (b) Beta diversity comparisons between patients and controls in both proliferative and secretory phases.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"2.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-7016822/v1/77f01729c5db3a29782e4dfc.png\"},{\"id\":86656104,\"identity\":\"1cbcf222-a4bc-4118-bafd-e006602616fa\",\"added_by\":\"auto\",\"created_at\":\"2025-07-14 10:17:30\",\"extension\":\"png\",\"order_by\":3,\"title\":\"Figure 3\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":115217,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003e(a) Genus-level relative abundance and microbial community profiles in proliferative phase samples. (b) Genus-level relative abundance and microbial community profiles in secretory phase samples.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"3.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-7016822/v1/4c86f1e42ffd75def13903d6.png\"},{\"id\":86660059,\"identity\":\"76593b24-f30a-4e4a-bf69-04802dbeb5ae\",\"added_by\":\"auto\",\"created_at\":\"2025-07-14 10:33:30\",\"extension\":\"png\",\"order_by\":4,\"title\":\"Figure 4\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":144185,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003e(a) Regression coefficients of differential taxa identified by MaAsLin2 (p ≤ 0.05) that distinguish endometriosis from controls in proliferative and secretory phase samples. (b) Boxplots illustrating the relative abundance distribution of differential taxa in proliferative and secretory phases. (c) Bacterial taxa identified as informative features through machine learning-based selection using the Random Forest algorithm.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"4.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-7016822/v1/aa6461ad9a662088a77ffe38.png\"},{\"id\":86656110,\"identity\":\"ae71bbf0-e53d-4a67-843e-660588837ef7\",\"added_by\":\"auto\",\"created_at\":\"2025-07-14 10:17:30\",\"extension\":\"png\",\"order_by\":5,\"title\":\"Figure 5\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":66757,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003e(a) Predictive performance of the differential microbial profile from the proliferative phase in classifying endometriosis. (b) Predictive performance of the differential microbial profile from the secretory phase in classifying endometriosis.\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"5.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-7016822/v1/aa36a4d701958aae1d4ec271.png\"},{\"id\":86660069,\"identity\":\"f0f077be-0d8f-489c-be63-97a4ad96eaf9\",\"added_by\":\"auto\",\"created_at\":\"2025-07-14 10:33:35\",\"extension\":\"pdf\",\"order_by\":0,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"manuscript-pdf\",\"size\":890415,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"manuscript.pdf\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-7016822/v1/731183b7-74b7-4c94-8383-f490fdb48946.pdf\"},{\"id\":86656102,\"identity\":\"cad58511-886a-4598-8359-baa6989b5177\",\"added_by\":\"auto\",\"created_at\":\"2025-07-14 10:17:30\",\"extension\":\"xlsx\",\"order_by\":0,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"supplement\",\"size\":56373,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"SupplementaryMaterials.xlsx\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-7016822/v1/6a82fe449da8547c87424807.xlsx\"}],\"financialInterests\":\"Competing interest reported. All authors, except Zhang Xinmei and Zhu Libo, are employees of Heranova Lifesciences, a\\ncompany engaged in the commercial development of a non-invasive test for endometriosis. Farideh Bischoff holds stock options of Heranova Lifesciences. The authors declare no other conflicts of interest, financial or otherwise.\",\"formattedTitle\":\"Uterine microbiome signatures associated with endometriosis\",\"fulltext\":[{\"header\":\"Background\",\"content\":\"\\u003cp\\u003eEndometriosis is a chronic inflammatory condition where endometrial-like tissue grows outside the uterus, often affecting the ovaries, fallopian tubes, and peritoneum (\\u003cspan citationid=\\\"CR1\\\" class=\\\"CitationRef\\\"\\u003e1\\u003c/span\\u003e, \\u003cspan citationid=\\\"CR2\\\" class=\\\"CitationRef\\\"\\u003e2\\u003c/span\\u003e). This ectopic tissue responds to hormonal changes, leading to symptoms like chronic pelvic pain, dysmenorrhea, dyspareunia, and infertility (\\u003cspan citationid=\\\"CR1\\\" class=\\\"CitationRef\\\"\\u003e1\\u003c/span\\u003e\\u0026ndash;\\u003cspan citationid=\\\"CR2\\\" class=\\\"CitationRef\\\"\\u003e2\\u003c/span\\u003e). Affecting an estimated 10% of reproductive-age women, the disease is frequently underdiagnosed, with an average delay of 6\\u0026ndash;11 years (\\u003cspan citationid=\\\"CR3\\\" class=\\\"CitationRef\\\"\\u003e3\\u003c/span\\u003e). This is due to its non-specific symptoms, overlap with other conditions, and the lack of reliable diagnostics.\\u003c/p\\u003e\\u003cp\\u003eMultiple theories exist about the origin of endometriosis \\u0026ndash; including retrograde menstruation, coelomic metaplasia, and stem cell involvement (\\u003cspan citationid=\\\"CR4\\\" class=\\\"CitationRef\\\"\\u003e4\\u003c/span\\u003e). Immune and inflammatory responses are also thought to play a key role, and recent research has begun exploring the gut and vaginal microbiomes as potential contributors (\\u003cspan citationid=\\\"CR5\\\" class=\\\"CitationRef\\\"\\u003e5\\u003c/span\\u003e). Studies show that women with endometriosis often have higher levels of \\u003cem\\u003eEscherichia coli\\u003c/em\\u003e in the gut, linked to elevated serum lipopolysaccharide, which activates inflammatory pathways and promotes pro-inflammatory cytokine production. These cytokines may aid the survival and implantation of ectopic endometrial cells (\\u003cspan citationid=\\\"CR6\\\" class=\\\"CitationRef\\\"\\u003e6\\u003c/span\\u003e). Additionally. \\u003cem\\u003eStreptococcus\\u003c/em\\u003e species have been linked to advanced disease stages, possibly by stimulating prostaglandin E2 expression, a contributor to pelvic pain (\\u003cspan citationid=\\\"CR7\\\" class=\\\"CitationRef\\\"\\u003e7\\u003c/span\\u003e). However, findings across studies are inconsistent; notably, one large study involving around 1000 participants found no significant association between gut microbiome and endometriosis (\\u003cspan citationid=\\\"CR8\\\" class=\\\"CitationRef\\\"\\u003e8\\u003c/span\\u003e). The vaginal microbiome, crucial for reproductive health, remains relatively understudied in the context of endometriosis (\\u003cspan citationid=\\\"CR9\\\" class=\\\"CitationRef\\\"\\u003e9\\u003c/span\\u003e). Shifts in vaginal microbiome, particularly the reduction of \\u003cem\\u003eLactobacillus\\u003c/em\\u003e and overgrowth of bacteria such as \\u003cem\\u003eGardnerella\\u003c/em\\u003e, \\u003cem\\u003ePrevotella\\u003c/em\\u003e and \\u003cem\\u003eMobiluncus\\u003c/em\\u003e are seen in bacterial vaginosis (BV), which is associated with inflammation and may contribute to endometriosis pathogenesis (\\u003cspan citationid=\\\"CR10\\\" class=\\\"CitationRef\\\"\\u003e10\\u003c/span\\u003e). Moreover, menstrual cycle phases influence vaginal microbial composition, with reduced \\u003cem\\u003eLactobacillus\\u003c/em\\u003e and increased diversity during menstruation, complicating research and emphasizing the importance of cycle-aware sampling (\\u003cspan citationid=\\\"CR11\\\" class=\\\"CitationRef\\\"\\u003e11\\u003c/span\\u003e).\\u003c/p\\u003e\\u003cp\\u003eTo investigate this further, we analyzed uterine microbiomes in 266 samples from women in either the proliferative or secretory phases. A total of 138 uterine tissue samples were collected from women in the proliferative phase of their menstrual cycle. Among these, 78 samples were obtained from women with a laparoscopic diagnosis of endometriosis, while the remaining 60 were from women without the disease. An additional 128 uterine tissue samples were collected during the secretory phase, including 88 from women diagnosed with endometriosis and 40 from unaffected individuals (Table\\u0026nbsp;\\u003cspan refid=\\\"Tab1\\\" class=\\\"InternalRef\\\"\\u003e1\\u003c/span\\u003e; Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig1\\\" class=\\\"InternalRef\\\"\\u003e1\\u003c/span\\u003e). Total genomic DNA was extracted from all tissue samples and used to prepare targeted bacterial 16S rRNA gene libraries for sequencing, as detailed in the Methods section. Raw sequence data were processed to remove technical artifacts, host DNA, and environmental contaminants. High-confidence bacterial reads were then taxonomically annotated using an internally curated version of the Greengenes2 database. Downstream analyses focused on identifying differentially abundant microbial taxa associated with endometriosis and determining taxa with potential predictive value for disease diagnosis.\\u003c/p\\u003e\\u003cp\\u003e\\u003cdiv class=\\\"gridtable\\\"\\u003e\\u003ctable float=\\\"Yes\\\" id=\\\"Tab1\\\" border=\\\"1\\\"\\u003e\\u003ccaption language=\\\"En\\\"\\u003e\\u003cdiv class=\\\"CaptionNumber\\\"\\u003eTable 1\\u003c/div\\u003e\\u003cdiv class=\\\"CaptionContent\\\"\\u003e\\u003cp\\u003eSummary of study samples. Median values are presented for each demographic parameter, with ranges shown in brackets. Statistical comparisons were conducted using T-tests.\\u003c/p\\u003e\\u003c/div\\u003e\\u003c/caption\\u003e\\u003ccolgroup cols=\\\"4\\\"\\u003e\\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c1\\\" colnum=\\\"1\\\"\\u003e\\u003c/div\\u003e\\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c2\\\" colnum=\\\"2\\\"\\u003e\\u003c/div\\u003e\\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c3\\\" colnum=\\\"3\\\"\\u003e\\u003c/div\\u003e\\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c4\\\" colnum=\\\"4\\\"\\u003e\\u003c/div\\u003e\\u003cthead\\u003e\\u003ctr\\u003e\\u003cth align=\\\"left\\\" colname=\\\"c1\\\"\\u003e\\u003cp\\u003e1. Characteristics\\u003c/p\\u003e\\u003c/th\\u003e\\u003cth align=\\\"left\\\" colname=\\\"c2\\\"\\u003e\\u003cp\\u003eEndometriosis\\u003c/p\\u003e\\u003c/th\\u003e\\u003cth align=\\\"left\\\" colname=\\\"c3\\\"\\u003e\\u003cp\\u003eControl\\u003c/p\\u003e\\u003c/th\\u003e\\u003cth align=\\\"left\\\" colname=\\\"c4\\\"\\u003e\\u003cp\\u003e\\u003cem\\u003ep\\u003c/em\\u003e-value\\u003c/p\\u003e\\u003c/th\\u003e\\u003c/tr\\u003e\\u003c/thead\\u003e\\u003ctbody\\u003e\\u003ctr\\u003e\\u003ctd align=\\\"left\\\" colspan=\\\"4\\\" nameend=\\\"c4\\\" namest=\\\"c1\\\"\\u003e\\u003cp\\u003e\\u003cem\\u003eAll samples (proliferative and secretory phases)\\u003c/em\\u003e\\u003c/p\\u003e\\u003c/td\\u003e\\u003c/tr\\u003e\\u003ctr\\u003e\\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e\\u003cp\\u003eSample size\\u003c/p\\u003e\\u003c/td\\u003e\\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e\\u003cp\\u003e166\\u003c/p\\u003e\\u003c/td\\u003e\\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e\\u003cp\\u003e100\\u003c/p\\u003e\\u003c/td\\u003e\\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e\\u003cp\\u003e-\\u003c/p\\u003e\\u003c/td\\u003e\\u003c/tr\\u003e\\u003ctr\\u003e\\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e\\u003cp\\u003eBMI\\u003c/p\\u003e\\u003c/td\\u003e\\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e\\u003cp\\u003e21.48 (15.62\\u0026ndash;36.85)\\u003c/p\\u003e\\u003c/td\\u003e\\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e\\u003cp\\u003e22.50 (17.1\\u0026ndash;31.22)\\u003c/p\\u003e\\u003c/td\\u003e\\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e\\u003cp\\u003e0.83\\u003c/p\\u003e\\u003c/td\\u003e\\u003c/tr\\u003e\\u003ctr\\u003e\\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e\\u003cp\\u003eAge\\u003c/p\\u003e\\u003c/td\\u003e\\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e\\u003cp\\u003e35.5 (20\\u0026ndash;51)\\u003c/p\\u003e\\u003c/td\\u003e\\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e\\u003cp\\u003e38.5 (21\\u0026ndash;50)\\u003c/p\\u003e\\u003c/td\\u003e\\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e\\u003cp\\u003e0.0002\\u003c/p\\u003e\\u003c/td\\u003e\\u003c/tr\\u003e\\u003ctr\\u003e\\u003ctd align=\\\"left\\\" colspan=\\\"4\\\" nameend=\\\"c4\\\" namest=\\\"c1\\\"\\u003e\\u003cp\\u003e\\u003cem\\u003eProliferative phase samples\\u003c/em\\u003e\\u003c/p\\u003e\\u003c/td\\u003e\\u003c/tr\\u003e\\u003ctr\\u003e\\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e\\u003cp\\u003eSample size\\u003c/p\\u003e\\u003c/td\\u003e\\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e\\u003cp\\u003e78\\u003c/p\\u003e\\u003c/td\\u003e\\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e\\u003cp\\u003e60\\u003c/p\\u003e\\u003c/td\\u003e\\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e\\u003cp\\u003e-\\u003c/p\\u003e\\u003c/td\\u003e\\u003c/tr\\u003e\\u003ctr\\u003e\\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e\\u003cp\\u003eBMI\\u003c/p\\u003e\\u003c/td\\u003e\\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e\\u003cp\\u003e21.51 (16.21\\u0026ndash;36.85)\\u003c/p\\u003e\\u003c/td\\u003e\\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e\\u003cp\\u003e22.50 (17.1\\u0026ndash;30.42)\\u003c/p\\u003e\\u003c/td\\u003e\\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e\\u003cp\\u003e0.76\\u003c/p\\u003e\\u003c/td\\u003e\\u003c/tr\\u003e\\u003ctr\\u003e\\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e\\u003cp\\u003eAge\\u003c/p\\u003e\\u003c/td\\u003e\\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e\\u003cp\\u003e36 (20\\u0026ndash;51)\\u003c/p\\u003e\\u003c/td\\u003e\\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e\\u003cp\\u003e40.5 (24\\u0026ndash;50)\\u003c/p\\u003e\\u003c/td\\u003e\\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e\\u003cp\\u003e0.026\\u003c/p\\u003e\\u003c/td\\u003e\\u003c/tr\\u003e\\u003ctr\\u003e\\u003ctd align=\\\"left\\\" colspan=\\\"4\\\" nameend=\\\"c4\\\" namest=\\\"c1\\\"\\u003e\\u003cp\\u003e\\u003cem\\u003eSecretory phase samples\\u003c/em\\u003e\\u003c/p\\u003e\\u003c/td\\u003e\\u003c/tr\\u003e\\u003ctr\\u003e\\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e\\u003cp\\u003eSample size\\u003c/p\\u003e\\u003c/td\\u003e\\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e\\u003cp\\u003e88\\u003c/p\\u003e\\u003c/td\\u003e\\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e\\u003cp\\u003e40\\u003c/p\\u003e\\u003c/td\\u003e\\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e\\u003cp\\u003e-\\u003c/p\\u003e\\u003c/td\\u003e\\u003c/tr\\u003e\\u003ctr\\u003e\\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e\\u003cp\\u003eBMI\\u003c/p\\u003e\\u003c/td\\u003e\\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e\\u003cp\\u003e21.45 (15.62\\u0026ndash;34.22)\\u003c/p\\u003e\\u003c/td\\u003e\\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e\\u003cp\\u003e22.49 (18.22\\u0026ndash;31.22)\\u003c/p\\u003e\\u003c/td\\u003e\\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e\\u003cp\\u003e0.94\\u003c/p\\u003e\\u003c/td\\u003e\\u003c/tr\\u003e\\u003ctr\\u003e\\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e\\u003cp\\u003eAge\\u003c/p\\u003e\\u003c/td\\u003e\\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e\\u003cp\\u003e35 (21\\u0026ndash;51)\\u003c/p\\u003e\\u003c/td\\u003e\\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e\\u003cp\\u003e37.5 (21\\u0026ndash;49)\\u003c/p\\u003e\\u003c/td\\u003e\\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e\\u003cp\\u003e0.005\\u003c/p\\u003e\\u003c/td\\u003e\\u003c/tr\\u003e\\u003c/tbody\\u003e\\u003c/colgroup\\u003e\\u003c/table\\u003e\\u003c/div\\u003e\\u003c/p\\u003e\\u003cp\\u003e\\u003c/p\\u003e\"},{\"header\":\"Results\",\"content\":\"\\u003cp\\u003e\\u003cem\\u003eUterine microbiome landscape in the study cohort\\u003c/em\\u003e\\u003c/p\\u003e\\u003cp\\u003eThe initial objective of our analysis was to evaluate both alpha and beta diversity of the uterine microbiome in women diagnosed with endometriosis compared to those without the disease, stratified by the proliferative and secretory phases of the menstrual cycle. Alpha diversity, which reflects the richness and evenness of microbial species within individual samples, was assessed using the Shannon Index. No statistically significant differences in alpha diversity were observed between endometriosis and control groups in either the proliferative or secretory phase (Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig2\\\" class=\\\"InternalRef\\\"\\u003e2\\u003c/span\\u003ea). Similarly, beta diversity, which measures compositional differences in microbial communities between groups, was evaluated using the Bray-Curtis dissimilarity metric. This analysis also revealed no significant differences between women with and without endometriosis across both menstrual phases (Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig2\\\" class=\\\"InternalRef\\\"\\u003e2\\u003c/span\\u003eb). These findings suggest that the overall diversity, including both the number of microbial taxa and their relative abundance distribution, is comparable between affected and unaffected individuals, irrespective of the menstrual cycle phase.\\u003c/p\\u003e\\u003cp\\u003e\\u003c/p\\u003e\\u003cp\\u003eGenus-level analysis revealed substantial variability in the relative abundance of \\u003cem\\u003eLactobacillus\\u003c/em\\u003e among individuals, both in patients and controls, across both menstrual phases. Although \\u003cem\\u003eLactobacillus\\u003c/em\\u003e is typically considered a hallmark of a healthy vaginal microbiome, its levels varied considerably, especially among patients with endometriosis in the proliferative phase compared to the controls (Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig3\\\" class=\\\"InternalRef\\\"\\u003e3\\u003c/span\\u003ea-b), recapitulating a previous study which showed considerable variations in \\u003cem\\u003eLactobacillus\\u003c/em\\u003e abundance even among healthy females (\\u003cspan citationid=\\\"CR11\\\" class=\\\"CitationRef\\\"\\u003e11\\u003c/span\\u003e). An analysis of the top 20 most abundant genera in both proliferative and secretory phase samples showed that the overall distribution of relative abundance was similar between patients and controls, with most of these dominant taxa not differentially abundant. One notable observation was made: the genus \\u003cem\\u003ePrevotella\\u003c/em\\u003e showed a trend toward enrichment in patient samples from the proliferative phase, with a borderline significant p-value (0.0509). Bacteria species in the genus \\u003cem\\u003ePrevotella\\u003c/em\\u003e are commonly associated with vaginal dysbiosis and pro-inflammatory states (\\u003cspan citationid=\\\"CR10\\\" class=\\\"CitationRef\\\"\\u003e10\\u003c/span\\u003e).\\u003c/p\\u003e\\u003cp\\u003e\\u003c/p\\u003e\\u003cp\\u003e\\u003cem\\u003eDifferentially abundant and machine-learning informative taxa\\u003c/em\\u003e\\u003c/p\\u003e\\u003cp\\u003eA more in-depth analysis of sub-genus level taxonomic units revealed eight taxa that were differentially abundant (p-value\\u0026thinsp;\\u0026le;\\u0026thinsp;0.05) between patients and controls in the proliferative phase, and three differential taxa in the secretory phase (Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig4\\\" class=\\\"InternalRef\\\"\\u003e4\\u003c/span\\u003ea), after adjusting for potential confounding effects of BMI and age. There is no overlap in differential taxa between proliferative and secretory phases. Although the number of differentially abundant taxa was modest, this finding aligns with expectations based on a prior large-scale study where it reported no statistically significant differences in gut microbial composition between women with and without endometriosis (\\u003cspan citationid=\\\"CR8\\\" class=\\\"CitationRef\\\"\\u003e8\\u003c/span\\u003e), despite earlier, smaller studies suggesting such associations. This discrepancy highlights the challenges in identifying consistent microbial biomarkers of gynecological disease, particularly in extraintestinal sites. Motivated by the hypothesis that the uterine microbiome may more directly reflect gynecological pathophysiology than the gut microbiome, we undertook this study to examine microbial community profiles in uterine tissue. In fact, after correcting for multiple comparisons using false discovery rate (FDR) adjustment, the initially observed differential taxa no longer reached statistical significance (i.e. FDR\\u0026thinsp;\\u0026gt;\\u0026thinsp;0.05). Nonetheless, we recognize that subtle yet consistent shifts in microbial composition across multiple taxa may carry predictive value (\\u003cspan citationid=\\\"CR12\\\" class=\\\"CitationRef\\\"\\u003e12\\u003c/span\\u003e). Therefore, we aim to employ machine learning approaches to integrate these signals, under the premise that the cumulative effect of multiple weakly informative taxa would enable and/or enhance predictive performance in distinguishing disease states (\\u003cspan citationid=\\\"CR13\\\" class=\\\"CitationRef\\\"\\u003e13\\u003c/span\\u003e). For instance, taxa such as \\u003cem\\u003ePrevotella\\u003c/em\\u003e sp.1 and \\u003cem\\u003eUreaplasma\\u003c/em\\u003e sp.1 \\u0026ndash; both belonging to genera frequently linked to bacterial vaginosis (\\u003cspan citationid=\\\"CR10\\\" class=\\\"CitationRef\\\"\\u003e10\\u003c/span\\u003e), while not individually conclusive after FDR correction, may collectively contribute to distinguishing disease states when incorporated as features in a supervised machine learning model.\\u003c/p\\u003e\\u003cp\\u003e\\u003c/p\\u003e\\u003cp\\u003eTo develop the feature set for supervised machine learning classification, we implemented a three-step selection strategy combining statistical and algorithmic criteria. First, we identified weakly differential taxa (i.e. nominal p-values\\u0026thinsp;\\u0026le;\\u0026thinsp;0.05; Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig4\\\" class=\\\"InternalRef\\\"\\u003e4\\u003c/span\\u003ea \\u0026amp; \\u003cspan refid=\\\"Fig4\\\" class=\\\"InternalRef\\\"\\u003e4\\u003c/span\\u003eb), indicating potential biological relevance despite not meeting strict multiple-testing thresholds. These taxa were initially included to ensure that subtle, non-random differences were not overlooked. In the second step, we applied a machine learning-based feature selection process by systematically evaluating the importance of each taxon detected in our profiling pipeline. This involved training preliminary models to score each taxon\\u0026rsquo;s contribution to classification performance, using a predefined threshold of feature importance score\\u0026thinsp;\\u0026ge;\\u0026thinsp;0.015 as a cutoff for inclusion. Taxa meeting this threshold were selected as additional candidates for the final feature set (Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig4\\\" class=\\\"InternalRef\\\"\\u003e4\\u003c/span\\u003ec; Supplementary Table \\u003cspan refid=\\\"MOESM1\\\" class=\\\"InternalRef\\\"\\u003eS1\\u003c/span\\u003e-S2). A subset of taxa was identified exclusively through the feature importance criterion. Specifically, 14 additional taxa including two other \\u003cem\\u003ePrevotella\\u003c/em\\u003e spp. were added in the proliferative phase. In the secretory phase, 11 additional taxa were incorporated. Notably, \\u003cem\\u003eGardnerella\\u003c/em\\u003e sp.1, a taxon traditionally linked to bacterial vaginosis was included in both proliferative and secretory cohorts. In the third step, a functional dysbiosis score (FDS) was calculated for each sample (Supplementary Table S3), representing an aggregate measure of microbial dysbiosis within the uterine tissue. The final feature set used for model training therefore comprised three components: the weakly differential taxa, taxa selected by feature importance scoring, and the FDS (Supplementary Table S4-S5).\\u003c/p\\u003e\\u003cp\\u003e\\u003cem\\u003eDiagnostic performance of machine-learning models in endometriosis detection\\u003c/em\\u003e\\u003c/p\\u003e\\u003cp\\u003eUsing the microbial profiles from the proliferative phase, we achieved moderate predictive performance in distinguishing endometriosis patients from controls. Specifically, across 50 rounds of repeated random subsampling cross-validation, the average AUC reached 0.70, indicating reasonable discriminative capability. The model demonstrated a sensitivity of 0.71, and a specificity of 0.54 (Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig2\\\" class=\\\"InternalRef\\\"\\u003e2\\u003c/span\\u003ed). While not optimal, the overall performance suggests that the microbiome during the proliferative phase carries meaningful signals that could aid in endometriosis diagnosis, especially when combined with other tools such as laparoscopy and blood-based biomarker tests. In contrast, models trained on the microbial profiles from secretory phase showed weaker overall performance, with an average AUC of only 0.58 (Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig5\\\" class=\\\"InternalRef\\\"\\u003e5\\u003c/span\\u003e). Overall, these findings suggest that microbial signatures differ in both menstrual phases, and that the proliferative phase carries more informative profiles. From a clinical and biological perspective, these results underscore the importance of menstrual cycle timing when considering the microbiome as a diagnostic aid for endometriosis. The result also suggests that cycle-phase-specific sampling may be crucial for optimizing microbiome-based diagnostics, and that future models may benefit from integrating hormonal phase information, or adjusting for it explicitly.\\u003c/p\\u003e\\u003cp\\u003e\\u003c/p\\u003e\"},{\"header\":\"Discussion\",\"content\":\"\\u003cp\\u003eThis study investigated the uterine microbiome in women with and without endometriosis, with a focus on menstrual cycle phase-specific microbial signatures and their predictive value for disease status. The differentially abundant taxa, coupled with the use of supervised machine-learning allowed us to uncover subtle patterns of microbial variation that may hold diagnostic potential.\\u003c/p\\u003e\\u003cp\\u003eOur findings add to a growing body of literature suggesting that microbial dysbiosis may contribute to endometriosis pathogenesis, though not through overt community-level disruption. Previous studies have implicated the gut and vaginal microbiomes, such as the elevated abundance of Gardnerella and Streptococcus being associated with advanced disease. However, large-scale studies such as the one by Perez-Prieto et al (2024), have failed to find significant associations between gut microbiome and endometriosis despite other studies suggesting so, indicating that location-specific microbial assessments may be more informative. Our results demonstrate that the uterine microbiome may contain weak but biologically meaningful signals, especially during the proliferative phase, where the tissue is more hormonally responsive and immunologically active. The modest predictive power (AUC\\u0026thinsp;=\\u0026thinsp;0.7) achieved by our model trained on proliferative phase cohort is consistent with recent studies showing that cumulative patterns of multiple taxa can outperform single-biomarker approaches. Furthermore, the menstrual cycle phase significantly influences reproductive tract microbiota, with menstruation being associated with reduced \\u003cem\\u003eLactobacillus\\u003c/em\\u003e and increased microbial diversity. These fluctuations likely contribute to inconsistencies across studies in the literature and emphasize the importance of phase-aware sampling. Our observation that microbial profiles from the secretory phase were less predictive supports this notion and aligns with reports that progesterone-dominant conditions may dampen inflammatory signals or microbiome-host interactions.\\u003c/p\\u003e\"},{\"header\":\"Conclusion\",\"content\":\"\\u003cp\\u003eIn conclusion, while the uterine microbiome alone is unlikely to serve as a stand-alone biomarker for endometriosis, our findings suggest that when analyzed in cycle-phase-aware and integrative manner, it may contribute to a broader diagnostic framework. Future studies should explore longitudinal sampling, integrate host transcriptomic and immunologic data, assess signals in vaginal mucus, and validate predictive models in larger, independent cohorts. Collectively, these efforts may pave the way toward a microbiome-informed, minimally invasive diagnostic tools for endometriosis.\\u003c/p\\u003e\"},{\"header\":\"Materials and Methods\",\"content\":\"\\u003cp\\u003e\\u003cem\\u003eSpecimen collection\\u003c/em\\u003e\\u003c/p\\u003e\\u003cp\\u003e This study was approved by the institutional review board of the Women\\u0026rsquo;s Hospital, Zhejiang University School of Medicine (IRB-20240110-R). Endometrial tissue samples were collected from 266 individuals, all of whom were clinically suspected of having a gynecologic condition and scheduled for laparoscopy with histopathological evaluation. To explore menstrual cycle-related differences, samples were obtained from women in either the proliferative or secretory phases of their cycle. The menstrual phase was initially assessed by physicians or surgeons based on self-reported cycle days and clinical evaluations. To confirm this classification, serum progesterone levels were measured using a protein assay from Kangrun Biotech Co. Ltd. (Guangdong, China), with levels above 1.08 ng/mL indicating the secretory phase, as per the manufacturer\\u0026rsquo;s guidelines. Among the samples, 138 were from the proliferative phase (78 from individuals with endometriosis and 60 from controls), and 128 were from the secretory phase (88 with endometriosis and 40 controls). Endometriosis was diagnosed and confirmed via gold-standard laparoscopic surgery. All participants provided written informed consent. The collected tissue samples were immediately transported at 4\\u0026deg;C to the Heranova Lifesciences laboratory and stored at -20\\u0026deg;C upon arrival.\\u003c/p\\u003e\\u003cp\\u003e\\u003cem\\u003eUterine tissue processing and targeted 16S library preparations\\u003c/em\\u003e\\u003c/p\\u003e\\u003cp\\u003eA 3\\u0026ndash;5 mm fragment of endometrial tissue from each sample was placed into an individual centrifuge tube with 20 \\u0026micro;L of Proteinase K and 180 \\u0026micro;L of Buffer ATL. The mixture was vortexed thoroughly and incubated at 58\\u0026deg;C with shaking at 1200 rpm for 3 hours. after incubation, add 210 \\u0026micro;L of Buffer ATL to each sample and homogeniz them using the TissueLyser II (2 minutes at 30 Hz, 1-minute pause, repeated for 15 cycles), and DNA was extracted using the QIAsymphony SP instrument (QIAGEN, 35459). DNA concentration and purity were assessed using the MultiSkan GO spectrophotometer (Thermo, 1510). The V4 variable region of the 16S rRNA was amplified using Invitrogen Platinum SuperFi II DNA Polymerase with the following PCR conditions: initial denaturation at 98\\u0026deg;C for 30 seconds (1 cycle), followed by 30 cycles of 98\\u0026deg;C for 10 seconds, 60\\u0026deg;C for 10 seconds, and 72\\u0026deg;C for 30 seconds, and a final extension at 72\\u0026deg;C for 5 minutes before holding at 4\\u0026deg;C. The forward primer is 5\\u0026rsquo;- TAATTGTGTGCCAGCMGCCGCGGTAA-3\\u0026rsquo; while the reverse primer is 5\\u0026rsquo;- TCAGCCGGACTACHVGGGTWTCTAAT-3\\u0026rsquo;. The PCR products were purified using VAHTS DNA Clean Beads. Adapter ligation was carried out using the UltraClean Universal DNA Library Prep Kit for Illumina V3 (Vazyme, UND607-02). First, 45 \\u0026micro;L of End Repair reaction mix was added to the purified PCR product, followed by incubation at 20\\u0026deg;C for 15 minutes, 65\\u0026deg;C for 15 minutes, and held at 4\\u0026deg;C. The ligation reaction mix was prepared on ice, added to the end-repaired DNA, and incubated at 20\\u0026deg;C for 15 minutes, then held at 4\\u0026deg;C. The ligated products were purified again with VAHTS DNA Clean Beads. Library amplification was performed under the following thermal conditions: 95\\u0026deg;C for 3 minutes (1 cycle), then 5 cycles of 98\\u0026deg;C for 20 seconds, 60\\u0026deg;C for 15 seconds, and 72\\u0026deg;C for 30 seconds, with a final extension at 72\\u0026deg;C for 5 minutes and a hold at 4\\u0026deg;C. The final libraries were purified using VAHTS DNA Clean Beads. Library concentrations were quantified using the KAPA Library Quantification Kit (KAPA, KK4824), and fragment sizes were evaluated with the Agilent 4200 TapeStation (Agilent, G2991A). Sequencing was performed on the Illumina MiSeq platform using the MiSeq Reagent Kit v2 (300 cycles).\\u003c/p\\u003e\\u003cp\\u003e\\u003cem\\u003eBioinformatic processing of targeted 16S sequencing data\\u003c/em\\u003e\\u003c/p\\u003e\\u003cp\\u003eThe demultiplexed FASTQ files from Illumina MiSeq sequencing were processed to extract the forward reads. To improve data quality, a two-step trimming and filtering process was employed. First, fastp (\\u003cspan citationid=\\\"CR14\\\" class=\\\"CitationRef\\\"\\u003e14\\u003c/span\\u003e) was used to identify and remove polyX artifacts\\u0026mdash;artificial stretches of a single nucleotide\\u0026mdash;commonly introduced during sequencing. Next, cutadapt (\\u003cspan citationid=\\\"CR15\\\" class=\\\"CitationRef\\\"\\u003e15\\u003c/span\\u003e) was used to trim any residual adapter sequences from the reads. To eliminate host-derived contamination, the filtered reads were aligned to the human reference genome (hg38) using Bowtie2 (\\u003cspan citationid=\\\"CR16\\\" class=\\\"CitationRef\\\"\\u003e16\\u003c/span\\u003e). Alignment results were processed with SAMtools (\\u003cspan citationid=\\\"CR17\\\" class=\\\"CitationRef\\\"\\u003e17\\u003c/span\\u003e), and reads mapping to the human genome were removed, ensuring that only non-host (primarily bacterial) sequences were retained for microbiome analysis.\\u003c/p\\u003e\\u003cp\\u003eThe resulting high-quality, non-human reads were then imported into the QIIME2 platform (\\u003cspan citationid=\\\"CR18\\\" class=\\\"CitationRef\\\"\\u003e18\\u003c/span\\u003e) for microbial community analysis. Within QIIME2, chimeric sequences were identified and removed using the vsearch uchime-denovo method. Subsequently, redundant sequences were collapsed using vsearch dereplicate-sequences, enhancing computational efficiency and reducing noise. The remaining high-confidence bacterial reads were annotated using an internally curated version of the Greengenes2 reference database (\\u003cspan citationid=\\\"CR19\\\" class=\\\"CitationRef\\\"\\u003e19\\u003c/span\\u003e). Taxonomic assignments were made using the Greengenes2 taxonomy-from-table classifier, providing genus-level and species-level annotations where possible. To ensure the validity and accuracy of the microbiome profiles, decontamination was performed using SCRuB (\\u003cspan citationid=\\\"CR20\\\" class=\\\"CitationRef\\\"\\u003e20\\u003c/span\\u003e), a statistical tool designed to identify and remove background contaminants. A blank negative control, which underwent the entire experimental workflow alongside the tissue samples, was included in the analysis to model and subtract any environmental or reagent-based contaminants. This approach ensured that the final dataset reflected true biological signals and minimized the risk of false microbial detection. Microbiome Shannon index and beta diversity were calculated and visualized using vegan (\\u003cspan citationid=\\\"CR21\\\" class=\\\"CitationRef\\\"\\u003e21\\u003c/span\\u003e) in R (v.4.4.2). A functional dysbiosis score was computed for each sample using the following formula (0.5*(1-Lactobacillus)\\u0026thinsp;+\\u0026thinsp;10*(Pathogenic taxa)) where pathogenic taxa consist of genus commonly associated bacterial vaginosis including \\u003cem\\u003eGardnerella\\u003c/em\\u003e, \\u003cem\\u003ePrevotella\\u003c/em\\u003e, \\u003cem\\u003eAnaerococcus\\u003c/em\\u003e, \\u003cem\\u003eStreptococcus\\u003c/em\\u003e, \\u003cem\\u003eMegasphaera\\u003c/em\\u003e, \\u003cem\\u003eMobiluncus\\u003c/em\\u003e, \\u003cem\\u003eSneathia\\u003c/em\\u003e, \\u003cem\\u003eAtopobium\\u003c/em\\u003e, \\u003cem\\u003ePeptoniphilus\\u003c/em\\u003e, \\u003cem\\u003eMycoplasmoides\\u003c/em\\u003e, \\u003cem\\u003eUreaplasma\\u003c/em\\u003e, \\u003cem\\u003eBacteroides\\u003c/em\\u003e, \\u003cem\\u003ePeptostreptococcus\\u003c/em\\u003e and \\u003cem\\u003eDialister\\u003c/em\\u003e.\\u003c/p\\u003e\\u003cp\\u003e\\u003cem\\u003eDisease prediction model construction using a random forest classifier\\u003c/em\\u003e\\u003c/p\\u003e\\u003cp\\u003e Samples were divided into proliferative (n\\u0026thinsp;=\\u0026thinsp;138) and secretory (n\\u0026thinsp;=\\u0026thinsp;128) groups for the following analysis, and the subsequent analysis was based on bacterial species relative abundances. MaAsLin2 (\\u003cspan citationid=\\\"CR22\\\" class=\\\"CitationRef\\\"\\u003e22\\u003c/span\\u003e) was performed to determine the multivariable association between bacterial species and endometriosis/non-endometriosis groups (\\u003cem\\u003eP\\u003c/em\\u003e\\u0026thinsp;\\u0026le;\\u0026thinsp;0.05), with age and BMI controlled using R package MaAsLin2. Features with importance scores\\u0026thinsp;\\u0026ge;\\u0026thinsp;0.015 to the endometriosis/non-endometriosis groups among bacterial species were selected via random forest implemented in Python sklearn package. Models to predict endometriosis/non-endometriosis were built based on features that were selected by MaAsLin2, by random forest feature scoring and by the addition of the functional dysbiosis score. Model performance was assessed through 50 iterations of repeated random subsampling cross-validation, in which the data was randomly split into 80% for training and 20% for testing in each iteration. This strategy helped account for variability arising from random data splits and yielded a more robust estimate of the predictive accuracy.\\u003c/p\\u003e\"},{\"header\":\"Declarations\",\"content\":\"\\u003cp\\u003e\\u003cem\\u003eEthics approval and consent to participate\\u003c/em\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eThis study was approved by the institutional review board of the Women\\u0026rsquo;s Hospital, Zhejiang University School of Medicine (IRB-20240110-R), and consent was given by enrolled participants.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cem\\u003eConsent for publication\\u003c/em\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eThe enrolled participants provided informed consent for the use of their biological materials in research and publication. All authors have reviewed and approved this manuscript for publication.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cem\\u003eData availability\\u003c/em\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eSequencing data were deposited into Genome Sequence Archive (accession number PRJCA041172)\\u0026nbsp;\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cem\\u003eConflict-of-interest statement\\u003c/em\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eAll authors, except Zhang Xinmei and Zhu Libo, are employees of Heranova Lifesciences, a\\u003c/p\\u003e\\n\\u003cp\\u003ecompany engaged in the commercial development of a non-invasive test for endometriosis. Farideh Bischoff holds stock options of Heranova Lifesciences. The authors declare no other conflicts of interest, financial or otherwise.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cem\\u003eFunding\\u003c/em\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eThe research was performed using internal R\\u0026amp;D budget at Heranova Lifesciences Inc.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cem\\u003eAuthor contributions\\u003c/em\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eL. Zhu and J. He analyzed the data. L. Zhu, J. He, W.H. Wong, F. Bischoff, X. Zhang conceptualized the study. L. Zhu, J. He and W.H. Wong wrote the manuscript with input from all authors. X. Zhang and L. Zhu provided the samples for sequencing. S. Lu coordinated sample collection and shipment to the laboratory. Y. Yu and X. Xu performed molecular experiments. W.H. Wong, F. Bischoff and X. Zhang supervised the study. F. Bischoff and X. Zhang secured R\\u0026amp;D funding for the study.\\u003c/p\\u003e\\n\\u003cp\\u003e\\u003cem\\u003eAcknowledgments\\u003c/em\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eThe authors would like to express their appreciation to Jonathan Zhao and Frank Zhang (both from Heranova Lifesciences) for their insightful input on the study design and manuscript.\\u003c/p\\u003e\"},{\"header\":\"References\",\"content\":\"\\u003col\\u003e\\n\\u003cli\\u003eZondervan KT, et al. Endometriosis. \\u003cem\\u003eNEJM\\u003c/em\\u003e. 2020;382:1244-1256\\u003c/li\\u003e\\n\\u003cli\\u003eParasar P, et al. Endometriosis: Epidemiology, diagnosis and clinical management. \\u003cem\\u003eCurr Obstet Gynecol Rep\\u003c/em\\u003e. 2017;6(1):34-41\\u003c/li\\u003e\\n\\u003cli\\u003eKirk UB, et al. Understanding endometriosis underfunding and its detrimental impact on awareness and research. \\u003cem\\u003enpj Women\\u0026rsquo;s Health\\u003c/em\\u003e. 2024;2:45\\u003c/li\\u003e\\n\\u003cli\\u003eLamceva J, et al. The main theories on the pathogenesis of endometriosis. \\u003cem\\u003eInternational Journal of Molecular Sciences\\u003c/em\\u003e. 2023;24(5):4254\\u003c/li\\u003e\\n\\u003cli\\u003eSer H, et al. Current updates on the role of microbiome in endometriosis: a narrative review. \\u003cem\\u003eMicroorganisms\\u003c/em\\u003e. 2023;11(2):360\\u003c/li\\u003e\\n\\u003cli\\u003eTavana Z, et al. Significant increased isolation of Escherichia coli in Iranian women with endometriosis: a case-control study. \\u003cem\\u003eBMC Women\\u0026rsquo;s Health\\u003c/em\\u003e. 2024;24:383\\u003c/li\\u003e\\n\\u003cli\\u003eShan J, et al. Gut microbiota imbalance and its correlations with hormone and inflammatory factors in patients with stage 3/4 endometriosis. \\u003cem\\u003eArch Gynecol Obstet\\u003c/em\\u003e. 2021;304(5):1363-1373\\u003c/li\\u003e\\n\\u003cli\\u003ePerez-Prieto I, et al. Gut microbiome in endometriosis: a cohort study on 1000 individuals. \\u003cem\\u003eBMC Medicine\\u003c/em\\u003e. 2024;22:294\\u003c/li\\u003e\\n\\u003cli\\u003eBrennan C, et al. Harnessing the power within: engineering the microbiome for enhanced gynecologic health. \\u003cem\\u003eReproduction and Fertility\\u003c/em\\u003e. 2024;5(2):e230060\\u003c/li\\u003e\\n\\u003cli\\u003eDing C, et al. Bacterial Vaginosis: effects on reproduction and its therapeutics. \\u003cem\\u003eJ Gynecol Obstet Hum Reprod\\u003c/em\\u003e. 2021;50(9):102174\\u003c/li\\u003e\\n\\u003cli\\u003eSong SD, et al. Daily vaginal microbiota fluctuations associated with natural hormonal cycle, contraceptives, diet and exercise. \\u003cem\\u003emSphere\\u003c/em\\u003e. 2020;5(4):e00593-20\\u003c/li\\u003e\\n\\u003cli\\u003eChang D, et al. Gut microbiome Wellness Index 2 enhances health status prediction from gut microbiome taxonomic profiles. \\u003cem\\u003eNature Communications\\u003c/em\\u003e. 2024;15:7447\\u003c/li\\u003e\\n\\u003cli\\u003eWang X, et al. Metagenomics reveals unique gut mycobiome biomarkers in major depressive disorder \\u0026ndash; a non-invasive method. \\u003cem\\u003eFrontiers in Cellular and Infection Microbiology\\u003c/em\\u003e. 2025;15:1582522\\u003c/li\\u003e\\n\\u003cli\\u003eChen S, et al. fastp: an ultra-fast all-in-one FASTQ preprocessor. \\u003cem\\u003eBioinformatics\\u003c/em\\u003e. 2018;34(17):i884-i890\\u003c/li\\u003e\\n\\u003cli\\u003eMarcel M. Cutadapt removes adapter sequences from high-throughput sequencing reads. \\u003cem\\u003eEMBnet.journal\\u003c/em\\u003e. 2011;17(1):10-12\\u003c/li\\u003e\\n\\u003cli\\u003eLangmead Ben \\u0026amp; Salzberg SL. Fast gapped-read alignment with Bowtie 2. \\u003cem\\u003eNature Methods\\u003c/em\\u003e. 2012;9:357-359\\u003c/li\\u003e\\n\\u003cli\\u003eLi H, et al. The sequence alignment/map format and SAMtools. \\u003cem\\u003eBioinformatics\\u003c/em\\u003e. 2009;25(16):2078-2079\\u003c/li\\u003e\\n\\u003cli\\u003eBolyen E, et al. Reproducible, interactive, scalable and extensible microbiome data science using QIIME 2. \\u003cem\\u003eNature Biotechnology\\u003c/em\\u003e. 2019;37:852-857\\u003c/li\\u003e\\n\\u003cli\\u003eMcDonald D, et al. Greengenes2 unifies microbial data in a single reference tree. \\u003cem\\u003eNature Biotechnology\\u003c/em\\u003e. 2024;42:715-718\\u003c/li\\u003e\\n\\u003cli\\u003eAustin GI, et al. Contamination source modeling with SCRuB improves cancer phenotype prediction from microbiome data. \\u003cem\\u003eNature Biotechnology\\u003c/em\\u003e. 2023;41:1820-1828\\u003c/li\\u003e\\n\\u003cli\\u003eDixon P. VEGAN, a package of R functions for community ecology. \\u003cem\\u003eJournal of Vegetation Science\\u003c/em\\u003e. 2003;14(6):927-930\\u003c/li\\u003e\\n\\u003cli\\u003eMallick H, et al. Multivariable association discovery in population-scale meta-omics studies. \\u003cem\\u003ePLoS Computational Biology\\u003c/em\\u003e. 2021;17(11):e100944\\u003c/li\\u003e\\n\\u003c/ol\\u003e\"}],\"fulltextSource\":\"\",\"fullText\":\"\",\"funders\":[],\"hasAdminPriorityOnWorkflow\":false,\"hasManuscriptDocX\":true,\"hasOptedInToPreprint\":true,\"hasPassedJournalQc\":\"\",\"hasAnyPriority\":false,\"hideJournal\":false,\"highlight\":\"\",\"institution\":\"\",\"isAcceptedByJournal\":true,\"isAuthorSuppliedPdf\":false,\"isDeskRejected\":\"\",\"isHiddenFromSearch\":false,\"isInQc\":false,\"isInWorkflow\":false,\"isPdf\":false,\"isPdfUpToDate\":true,\"isWithdrawnOrRetracted\":false,\"journal\":{\"display\":true,\"email\":\"info@researchsquare.com\",\"identity\":\"bmc-biology\",\"isNatureJournal\":false,\"hasQc\":true,\"allowDirectSubmit\":false,\"externalIdentity\":\"\",\"sideBox\":\"Learn more about [BMC Biology](https://bmcbiol.biomedcentral.com/)\",\"snPcode\":\"12915\",\"submissionUrl\":\"https://submission.springernature.com/new-submission/12915/3\",\"title\":\"BMC Biology\",\"twitterHandle\":\"\",\"acdcEnabled\":true,\"dfaEnabled\":true,\"editorialSystem\":\"stoa\",\"reportingPortfolio\":\"BMC Series\",\"inReviewEnabled\":true,\"inReviewRevisionsEnabled\":true},\"keywords\":\"\",\"lastPublishedDoi\":\"10.21203/rs.3.rs-7016822/v1\",\"lastPublishedDoiUrl\":\"https://doi.org/10.21203/rs.3.rs-7016822/v1\",\"license\":{\"name\":\"CC BY 4.0\",\"url\":\"https://creativecommons.org/licenses/by/4.0/\"},\"manuscriptAbstract\":\"\\u003ch2\\u003eBackground\\u003c/h2\\u003e\\u003cp\\u003eEndometriosis is a chronic inflammatory disorder affecting\\u0026thinsp;~\\u0026thinsp;10% of reproductive-age women, often causing pelvic pain and infertility. Despite its prevalence, diagnosis remains delayed due to non-specific symptoms and lack of reliable non-invasive biomarkers. Emerging evidence implicates the microbiome in disease pathogenesis.\\u003c/p\\u003e\\u003ch2\\u003eResults\\u003c/h2\\u003e\\u003cp\\u003eWe analyzed uterine microbiomes from 266 tissue samples collected during either the proliferative or secretory phase, using 16S rRNA gene sequencing. Genus-level analysis revealed variable \\u003cem\\u003eLactobacillus\\u003c/em\\u003e abundance among all individuals. \\u003cem\\u003ePrevotella\\u003c/em\\u003e showed borderline enrichment in proliferative-phase patients. Sub-genus analyses identified a small number of differentially abundant taxa, though none remained significant after FDR correction. To capture subtle microbial shifts, we developed a feature set combining weakly differential taxa, algorithmically selected taxa via machine-learning, and a functional dysbiosis score. A supervised classifier trained on proliferative-phase data achieved moderate predictive performance (AUC\\u0026thinsp;=\\u0026thinsp;0.70), while secretory-phase models performed poorly (AUC\\u0026thinsp;=\\u0026thinsp;0.58).\\u003c/p\\u003e\\u003ch2\\u003eConclusion\\u003c/h2\\u003e\\u003cp\\u003eThe uterine microbiome shows phase-dependent differences in its potential to inform endometriosis status. Although no robust individual microbial biomarkers were identified, machine-learning models incorporating subtle community features from the proliferative phase yielded modest diagnostic potential. These results highlight the importance of menstrual cycle-aware sampling and support further development of microbiome-informed diagnostic tools for endometriosis.\\u003c/p\\u003e\",\"manuscriptTitle\":\"Uterine microbiome signatures associated with endometriosis\",\"msid\":\"\",\"msnumber\":\"\",\"nonDraftVersions\":[{\"code\":1,\"date\":\"2025-07-14 10:17:25\",\"doi\":\"10.21203/rs.3.rs-7016822/v1\",\"editorialEvents\":[{\"type\":\"communityComments\",\"content\":0},{\"type\":\"decision\",\"content\":\"Revision requested\",\"date\":\"2025-12-19T20:36:41+00:00\",\"index\":\"\",\"fulltext\":\"\"},{\"type\":\"editorInvitedReview\",\"content\":\"\",\"date\":\"2025-12-14T13:40:43+00:00\",\"index\":\"hide\",\"fulltext\":\"\"},{\"type\":\"reviewerAgreed\",\"content\":\"290812013872811398632811092058393570953\",\"date\":\"2025-12-02T05:45:23+00:00\",\"index\":\"hide\",\"fulltext\":\"\"},{\"type\":\"editorInvitedReview\",\"content\":\"\",\"date\":\"2025-09-29T19:00:59+00:00\",\"index\":\"hide\",\"fulltext\":\"\"},{\"type\":\"reviewerAgreed\",\"content\":\"30595913225908239772456379392068839330\",\"date\":\"2025-08-29T07:37:07+00:00\",\"index\":\"hide\",\"fulltext\":\"\"},{\"type\":\"reviewerAgreed\",\"content\":\"164715783114060013993462450734103631356\",\"date\":\"2025-07-23T11:13:15+00:00\",\"index\":\"hide\",\"fulltext\":\"\"},{\"type\":\"reviewersInvited\",\"content\":\"\",\"date\":\"2025-07-08T13:24:35+00:00\",\"index\":\"\",\"fulltext\":\"\"},{\"type\":\"editorAssigned\",\"content\":\"\",\"date\":\"2025-07-01T12:55:11+00:00\",\"index\":\"\",\"fulltext\":\"\"},{\"type\":\"checksComplete\",\"content\":\"\",\"date\":\"2025-07-01T06:59:21+00:00\",\"index\":\"\",\"fulltext\":\"\"},{\"type\":\"submitted\",\"content\":\"BMC Biology\",\"date\":\"2025-07-01T06:40:11+00:00\",\"index\":\"\",\"fulltext\":\"\"}],\"status\":\"published\",\"journal\":{\"display\":true,\"email\":\"info@researchsquare.com\",\"identity\":\"bmc-biology\",\"isNatureJournal\":false,\"hasQc\":true,\"allowDirectSubmit\":false,\"externalIdentity\":\"\",\"sideBox\":\"Learn more about [BMC Biology](https://bmcbiol.biomedcentral.com/)\",\"snPcode\":\"12915\",\"submissionUrl\":\"https://submission.springernature.com/new-submission/12915/3\",\"title\":\"BMC Biology\",\"twitterHandle\":\"\",\"acdcEnabled\":true,\"dfaEnabled\":true,\"editorialSystem\":\"stoa\",\"reportingPortfolio\":\"BMC Series\",\"inReviewEnabled\":true,\"inReviewRevisionsEnabled\":true}}],\"origin\":\"\",\"ownerIdentity\":\"340d01ba-f0f9-4ec9-91b0-e6378d5ffd7a\",\"owner\":[],\"postedDate\":\"July 14th, 2025\",\"published\":true,\"recentEditorialEvents\":[],\"rejectedJournal\":[],\"revision\":\"\",\"amendment\":\"\",\"status\":\"under-review\",\"subjectAreas\":[],\"tags\":[],\"updatedAt\":\"2026-06-11T14:01:34+00:00\",\"versionOfRecord\":[],\"versionCreatedAt\":\"2025-07-14 10:17:25\",\"video\":\"\",\"vorDoi\":\"\",\"vorDoiUrl\":\"\",\"workflowStages\":[]},\"version\":\"v1\",\"identity\":\"rs-7016822\",\"journalConfig\":\"researchsquare\"},\"__N_SSP\":true},\"page\":\"/article/[identity]/[[...version]]\",\"query\":{\"redirect\":\"/article/rs-7016822\",\"identity\":\"rs-7016822\",\"version\":[\"v1\"]},\"buildId\":\"B-jG_2CBjPDmsCi4Wdhf-\",\"isFallback\":false,\"isExperimentalCompile\":false,\"dynamicIds\":[84888],\"gssp\":true,\"scriptLoader\":[]}","source_license":"CC0","license_restricted":false}