Machine learning–driven decoding of maternal immune signatures in repeated pregnancy loss

This article has 0 evaluations Published on
Read the full article Related papers
This article on Sciety

Abstract

Background

Repeated pregnancy loss (RPL) is a multifactorial condition in which the underlying immunological mechanisms remain incompletely understood. Although immune tolerance at the maternal–fetal interface is critical for successful pregnancy, the immune disruptions that contribute to RPL, independent of fetal aneuploidy, remain poorly characterized.

Methods

We performed single-cell RNA sequencing of decidual tissues from RPL patients and gestational age–matched controls. We employed single-cell transcriptomics coupled with genotype-based origin analysis to dissect immune dysregulation in RPL. To identify cell type-specific RPL signatures, we mutually applied a supervised machine learning model and foundation model-based analysis, followed by pathway enrichment and network analysis. In addition, we prioritized drug repurposing candidates based on the drug response data, which reversed the RPL-associated transcriptional profiles.

Results

In normal pregnancy, fetal immune cells were nearly absent but increased in RPL, whereas fetal trophoblasts were reduced, suggesting hindered placental development despite the absence of chromosomal abnormalities in fetal cells. Interestingly, by showing a strong common rejection module score, RPL immune cells resemble the transcriptional signatures of acute transplant rejection, indicating the contribution of the maternal immune system. Our machine learning and transformer-based models mutually identified T-cell–derived transcriptomic signatures that distinguished RPL immune cells. By examining biological confounding factors, including fetal-origin signatures, we prioritized CXCR4 and JUN in maternal T cells as RPL-associated signatures. For clinical application, we explored reversed transcriptional signatures from drug response data, and three compounds were highlighted as candidates for drug repurposing.

Conclusions

Together, our approach identifies maternal immune signatures such as those of CXCR4 and JUN and links them to potential drug repurposing candidates, thereby providing both mechanistic insights and therapeutic opportunities for RPL.

Related articles

Related articles are currently not available for this article.