Single-Cell Transcriptomic Profiling and Machine Learning Integration Unveil Stromal Cell Heterogeneity in...

Area of Science:

• Genomics and Molecular Biology
• Computational Biology and Bioinformatics
• Reproductive Medicine

Background:

• Endometriosis (EMs) affects 10% of reproductive-age women globally, with unclear pathogenesis.
• Abnormal cell differentiation and somatic mutations in ectopic endometrium are key to EMs progression and treatment variability.
• Understanding molecular mechanisms driving ectopic endometrial cell differentiation is crucial for EMs management.

Purpose of the Study:

• To elucidate molecular mechanisms of ectopic endometrial cell differentiation using machine learning (ML) and single-cell RNA sequencing (scRNA-seq).
• To identify novel prognostic biomarkers and therapeutic targets for EMs, focusing on mutation-driven transcriptional alterations.
• To stratify EMs patients and characterize the ectopic microenvironment.

Main Methods:

• Analysis of comprehensive transcriptomic data from GEO and HED, including scRNA-seq data from 162,485 cells across 46 EMs patients.
• Application of 10 ML algorithms and 101 hybrid combinations for predictive modeling and patient stratification.
• Unsupervised clustering, functional enrichment, pathway analysis, and cell-cell communication network construction.

Main Results:

• Identification of 298 genes associated with ectopic endometrial cell differentiation, including mutation-harboring genes.
• Discovery of two distinct patient subgroups (high- and low-invasive phenotypes) with different disease trajectories.
• Validation of differential expression for HOXA10 (downregulated), ESR1, MMP9, and SPP1 (upregulated) in ectopic vs. normal endometrial stromal cells.

Conclusions:

• Comprehensive molecular characterization of EMs cell differentiation via ML and scRNA-seq.
• Identification of distinct patient phenotypes, key regulatory genes (HOXA10, ESR1, MMP9, SPP1), and macrophage-centric communication networks.
• HOXA10, ESR1, MMP9, and SPP1 as potential diagnostic biomarkers and therapeutic targets for personalized EMs treatment.