Exploration of shared pathogenic factors and causative genes in early-stage endometrial cancer and osteoarthritis

Abstract

Osteoarthritis (OA) has been implicated in the development and progression of early-stage endometrial cancer (EC), suggesting shared pathogenic factors between the two diseases. This study aimed to investigate the causal relationship between OA and EC and to identify causative genes common to both early-stage EC and OA. A Two-sample Mendelian randomization (MR) analysis was first performed to assess the causal relationship between OA and EC. Differentially expressed genes associated with early-stage EC and OA were identified using the limma package. Overlapping genes were extracted to determine common causative genes, followed by enrichment analysis. The causal relationship between these genes and EC was verified through Mendelian randomization (MR) of drug targets. Genes with diagnostic value were identified using multiple machine learning algorithms to construct EC prediction models and evaluate their performance. Additionally, the study examined the correlation between diagnostic-value genes and immune cell infiltration. IVW analysis indicated that OA was a high-risk factor for the development of EC (P < 0.05). Seven common causative genes (CDKN2A, DDA1, LRRC42, POLB, ADCYAP1R1, DNMT3A, and GLRX5) were identified for OA and EC, showing significant enrichment in related pathways such as heterochromatin. MR analysis of drug targets revealed that CDKN2A, DDA1, LRRC42, and POLB had diagnostic value for EC. The EC prediction model based on these four genes demonstrated high performance (AUC = 0.974 for the training set; AUC = 0.966 for the validation set), and these genes were significantly associated with immune cell infiltration (P < 0.05). CDKN2A, DDA1, LRRC42, and POLB may be common causative genes for OA and early-stage EC, potentially serving as targets for drug intervention.

Keywords: Endometrial cancer; Enrichment analysis; Mendelian randomization; Nomogram; Osteoarthritis.

Fig. 1

Schematic diagram of MR associated with EC. Three major assumptions: ① The assumption of association: The instrumental variable is closely related to the exposure factor. ② The assumption of association: The instrumental variable is closely related to the exposure factor. ③ The assumption of independence: The instrumental variable is not correlated with confounders.

Fig. 2

Two-sample MR analysis of OA and EC. (A) Scatter plot (B) ‘Leave-one-out’ sensitivity analysis (C) Funnel plot.

Fig. 3

Differential genes for EC and OA. (A) Differentially expressed genes in Stage I EC tissues versus normal endometrial tissues. (B) Differentially expressed genes in Stage I EC tissues versus Stage II and III EC tissues. (C) Differentially expressed genes in OA tissues versus normal tissues. (D) Genes with differential expression in opposite directions compared to Fig. 5A and Fig. 5B. (E) Genes with the same direction of differential expression as in Figure C and Figure D. (F) GO enrichment analysis.

Fig. 4

Forest plot of the results of causal association analysis between differential gene eQTLs and EC.

Fig. 5

Screening of genes with diagnostic value using multiple machine learning algorithms. (A: Construction of an EC prediction model using LASSO modelling B: Construction of an EC prediction model using RF modelling, C: The two models A and B take the intersection.).

Fig. 6

Construction of the prediction model. (A) Nomogram for the 4 genes with diagnostic value (B) ROC training curves for the 4 genes with diagnostic value (C) ROC testing curves for the 4 genes with diagnostic value (D) Training curves for nomogram prediction models (E) Testing curves for nomogram prediction models (F) DCA training Curve for nomogram prediction models (G) DCA testing Curve for nomogram prediction models.

Fig. 7

Correlation analysis of immune cell infiltration (ns, p ≥ 0.05;*, p < 0.05;**, p < 0.01;***, p < 0.001).