# Construction and validation of a novel diagnostic model for esophageal squamous cell carcinoma: an integrated analysis of multi-omics data

## TL;DR

Researchers developed a new diagnostic model for esophageal squamous cell carcinoma using multi-omics data and identified five key genes for early detection.

## Contribution

A novel diagnostic model for ESCC using integrated multi-omics data and identification of five robust biomarkers.

## Key findings

- A diagnostic model with five genes achieved high accuracy (AUCs of 0.99 and 0.97) in training and validation sets.
- SORBS2 was identified as a tumor-suppressor gene predominantly expressed in myofibroblasts in ESCC tissue.
- Single-cell RNA analysis revealed myofibroblasts as the main source of SORBS2 expression in tumor tissue.

## Abstract

Esophageal squamous cell carcinoma (ESCC), highly prevalent in China, has a limited number of ideal genes for early diagnosis, highlighting the need for the development of novel biomarkers to improve detection capabilities. The purpose of this study is to develop and validate a new genetic diagnostic model for ESCC.

Publicly available bulk RNA-seq datasets (GSE23400, GSE17351, GSE20347) were merged to identify differentially expressed genes (DEGs) between ESCC and adjacent normal tissues. Weighted gene co-expression network analysis (WGCNA) and protein-protein interaction (PPI) were performed to identify hub genes associated with ESCC. We identified the intersecting genes between the DEGs and those within the ESCC-related module identified by WGCNA. We subsequently refined these intersecting genes via LASSO regression and then constructed a diagnostic model for ESCC using multivariate logistic regression. ESCC samples from the TCGA database were used as the external validation set. Validation of the identified protective factor was conducted through Western blotting (WB) in mouse ESCC models and immunofluorescence (IF) in human tissues. Additionally, single-cell RNA analysis was conducted to explore the cell types expressing the marker genes.

113 upregulated and 173 downregulated genes were found in the ESCC groups. WGCNA identified the blue module (13 genes) as most correlated with ESCC. We obtained a total of 13 intersecting genes. Among them, five genes formed the diagnostic model: Logit(P) = −24.4547 + 2.0567×BID + 0.7396×CBX3 + 2.3757×ECT2 + 0.5667×KIF14 − 2.1019×SORBS2. The model achieved AUCs of 0.99 (training set) and 0.97 (external validation set). SORBS2 was the only potential protective factor in the model. WB indicated higher expression levels of SORBS2 in the adjacent normal esophageal tissue compared to those in the ESCC tissue. single-cell RNA analysis revealed that myofibroblasts are the predominant cellular source of SORBS2 expression within ESCC tumor tissue. IF confirmed lower level of SORBS2 expression in myofibroblast in the ESCC than those in the adjacent normal esophageal tissue.

We developed an ESCC diagnostic model and identified BID, CBX3, ECT2, KIF14, and SORBS2 as robust ESCC biomarkers. SORBS2 is a tumor-suppressor gene predominantly expressed in myofibroblasts.