Scholarlink

<p>Mouse spermatogenesis is a highly orchestrated process, yet the transcriptional and molecular heterogeneity of spermatogonial stem cells (SSCs) remains incompletely understood. Leveraging multi-omics integration and machine learning, we aimed to characterize SSC subtypes, identify key regulatory genes, and uncover novel biomarkers associated with spermatogenic function and stemness. We systematically retrieved and curated four mouse gene expression datasets from GEO (GSE45887, GSE45885, GSE9210, and validation set GSE145467), applying rigorous preprocessing, normalization, and batch effect correction using R/Bioconductor tools. Differentially expressed genes (DEGs) were identified using the Limma package (FDR &lt; 0.01, |log₂FC| &gt; 1.5), followed by GO and KEGG enrichment analyses via Enrichr. Gene Set Variation Analysis (GSVA) and a stemness index (mRNAsi) based on the OCLR algorithm were computed. Key hub genes were validated using scRNA-seq data (GSE149512) and further assessed by immunohistochemistry (IHC) and biomarker selection via recursive feature elimination (RFE). Integration of multi-dataset transcriptomic profiles revealed a conserved set of DEGs involved in stemness, germ cell development, and metabolic regulation. Hierarchical clustering and GSVA distinguished distinct SSC subtypes, with enrichment in hallmark and KEGG pathways linked to glycolysis, cell cycle regulation, and pluripotency. Single-cell analyses confirmed expression of key SSC markers (e.g., CD9, ID4, THY1), while SIMPA-based imputation of single-cell ChIP-seq data validated epigenetic occupancy patterns. Machine learning models (Random Forest, SVM, ANN) achieved high classification performance (AUC &gt; 0.92) in predicting SSC phenotypes from transcriptomic and IHC data. IHC validation in mouse testicular tissues further corroborated spatial localization of hub genes. Our integrative transcriptomic and single-cell approach uncovers the molecular landscape of mouse SSCs, identifies robust biomarkers of spermatogenic function, and presents a computational pipeline for SSC characterization. These findings offer critical insights for reproductive medicine and potential therapeutic applications in male infertility and regenerative biology.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Identifying Mouse Undifferentiated To Differentiated Spermatogonia Stem Cells at the Single-Cell Level Using Machine Learning Approaches

  • 期刊论文

  • 发表日期：

    2026年2月13日

查看全文
  • Danial Hashemi Karoii,
  • Zahra Hasani Mahforoozmahalleh,
  • Maryam Osanloo,
  • Hossein Azizi,
  • Thomas Skutella

摘要

Mouse spermatogenesis is a highly orchestrated process, yet the transcriptional and molecular heterogeneity of spermatogonial stem cells (SSCs) remains incompletely understood. Leveraging multi-omics integration and machine learning, we aimed to characterize SSC subtypes, identify key regulatory genes, and uncover novel biomarkers associated with spermatogenic function and stemness. We systematically retrieved and curated four mouse gene expression datasets from GEO (GSE45887, GSE45885, GSE9210, and validation set GSE145467), applying rigorous preprocessing, normalization, and batch effect correction using R/Bioconductor tools. Differentially expressed genes (DEGs) were identified using the Limma package (FDR < 0.01, |log₂FC| > 1.5), followed by GO and KEGG enrichment analyses via Enrichr. Gene Set Variation Analysis (GSVA) and a stemness index (mRNAsi) based on the OCLR algorithm were computed. Key hub genes were validated using scRNA-seq data (GSE149512) and further assessed by immunohistochemistry (IHC) and biomarker selection via recursive feature elimination (RFE). Integration of multi-dataset transcriptomic profiles revealed a conserved set of DEGs involved in stemness, germ cell development, and metabolic regulation. Hierarchical clustering and GSVA distinguished distinct SSC subtypes, with enrichment in hallmark and KEGG pathways linked to glycolysis, cell cycle regulation, and pluripotency. Single-cell analyses confirmed expression of key SSC markers (e.g., CD9, ID4, THY1), while SIMPA-based imputation of single-cell ChIP-seq data validated epigenetic occupancy patterns. Machine learning models (Random Forest, SVM, ANN) achieved high classification performance (AUC > 0.92) in predicting SSC phenotypes from transcriptomic and IHC data. IHC validation in mouse testicular tissues further corroborated spatial localization of hub genes. Our integrative transcriptomic and single-cell approach uncovers the molecular landscape of mouse SSCs, identifies robust biomarkers of spermatogenic function, and presents a computational pipeline for SSC characterization. These findings offer critical insights for reproductive medicine and potential therapeutic applications in male infertility and regenerative biology.