# sMICA/sMICB and Immune Checkpoint in Endometriosis: Toward a Minimally Invasive Diagnostic Model Based on Machine Learning

## TL;DR

This study explores immune markers in endometriosis and proposes a machine learning model for a less invasive diagnosis.

## Contribution

A novel machine learning-based diagnostic model for endometriosis using immune markers and patient data.

## Key findings

- sMICB levels in peritoneal fluid differ by endometriosis stage and adhesion presence.
- sMICA levels in peritoneal fluid are elevated in infertility-related endometriosis.
- XGBoost model outperformed logistic regression with higher accuracy and AUC for endometriosis diagnosis.

## Abstract

Background: Endometriosis is a complex condition that impairs women’s quality of life and reproductive potential. Its diagnosis remains significant challenge for clinicians. The aim of the study was to investigate cancer-like immune evasion mechanisms in endometriosis and to develop a novel diagnostic model using machine learning. Methods: In this study, we measured the levels of soluble forms of the following immune markers in blood serum and peritoneal fluid (PF): sMICA, sMICB, sEng, sCD25, s4-1BB, sB7.2, sCTLA-4, sPD-L1, sPD-1, sTIM-3, sLAG-3, and sGal-9. Results: sMICB levels in PF differed across endometriosis stages and were higher in patients with endometriosis-associated adhesions. sMICA levels in PF were elevated in women with endometriosis-associated infertility. The disease severity was inversely correlated with serum sB7.2 levels and positively correlated with serum sTIM-3 levels. A logistic regression model achieved an accuracy = 0.79, AUC = 0.94, and F1-score = 0.88, whereas XGBoost performed better with accuracy = 0.94, AUC = 0.95, and F1-score = 0.96. The key predictive features in both models were sMICB serum level and patients’ pain score. Conclusions: Our results demonstrate the potential role of sMICA and sMICB shedding in endometriosis and present a novel, minimally invasive diagnostic approach.