# A lactylation modification-related prediction model for the diagnosis of ulcerative colitis based on machine learning

## TL;DR

This paper uses machine learning to identify lactylation-related genes that can help diagnose ulcerative colitis and explores their role in immune cells.

## Contribution

The study introduces a novel machine learning-based prediction model for UC diagnosis using lactylation-related genes.

## Key findings

- A lactylation-related prediction model with four core genes achieved high diagnostic accuracy (AUC: 0.976) for UC.
- HSC, NK, and macrophage cells showed higher lactylation-related scores in UC patients.
- Nala and 2-DG treatments significantly altered the expression of the four core lactylation-related genes.

## Abstract

Lactylation modification serves as a critical link between metabolic reprogramming and epigenetic regulation, playing a significant role in the progression of both malignant tumors and inflammatory diseases. Nevertheless, its specific function in the pathogenesis of ulcerative colitis (UC) remains poorly understood.

The hub genes associated with lactylation in UC were identified and validated by mining three UC-related datasets (GSE206285, GSE75214, and GSE87466) from the GEO database, and we created a lactylation-related prediction model for the diagnosis of UC. The lactylation levels of different immune cells were also investigated via single-cell (sc) RNA-sequencing data. Finally, the core genes of lactylation were validated in vitro.

Four lactylation-related core genes (HIF1A, SLC25A12, SLC16A3, and PFKFB2) that are closely correlated with UC were identified by three machine learning methods, and the lactylation-related prediction model based on the four genes exhibited outstanding diagnostic performance for UC (AUC:0.976, 95% CI: 0.941–1.00). scRNA-sequencing analysis revealed that HSC, NK, and macrophage cells exhibited higher lactylation-related scores in UC compared to other immune cells. After Nala intervention, the expressions of the four core genes were significantly increased, while the expressions of the four genes were significantly decreased after treatment with 2-DG.

By applying machine learning methods to analyze sequencing data, we identified core lactylation-related genes in UC and developed a diagnostic model with high predictive performance. Furthermore, based on scRNA-seq data, we investigated lactylation modifications across seven types of immune cells in UC patients, providing valuable insights into the interplay between lactylation and immune cells in UC.