Bioinformatics and Machine Learning-Based Identification of Critical Biomarkers and Immune Infiltration in Venous...

Objective: This study aims to use bioinformatics and machine learning algorithms to screen and analyze the key genes involved in venous thromboembolism (VTE) and explore the relationship between these biomarkers and immune cell infiltration. Methods: The gene expression profile with the identifier GSE19151 was downloaded from the GEO database. Differential expression analysis using the limma package was conducted to identify genes that were differentially expressed between VTE and normal samples. Biological activities of these genes were then investigated through GO analysis utilizing the R language package. KEGG and GSEA were also performed to identify key signaling pathways. Furthermore, machine learning techniques were employed to determine hub gene signatures related to VTE, and ROC curves were used to validate the findings. To compare the immune infiltration of healthy and VTE samples, single sample gene set enrichment analysis (ssGSEA) was applied. Lastly, the Spearman correlation coefficient was used to assess the relationship between the expression of hub genes and immune cell infiltration. Results: A total of 628 differentially expressed genes (DEGs) were discovered between the VTE samples and normal samples. GO analysis identified protein polyubiquitination, lysosomal lumen acidification, organellar ribosome, mitochondrial ribosome, ammonium transmembrane transporter activity, and immunoglobulin binding as the processes with the highest abundance of DEGs. KEGG pathway analysis revealed that DEGs were enriched in ribosome, COVID-19, viral infection, oxidative phosphorylation, Parkinson's disease, nonalcoholic fatty liver disease, apoptosis, and cancer. The most prominent KEGG pathways associated with VTE were ribosome, Parkinson's disease, oxidative phosphorylation, Alzheimer's disease, and Huntington's disease according to GSEA findings. DLST and LSP1 were identified as hub gene signatures in VTE by machine learning integrative analysis, and ROC curves confirmed their diagnostic value. Results from ssGSEA indicated a significant difference in the degree of immune cell infiltration between VTE and normal samples, with the expression of DLST and LSP1 positively correlated with the content of some immune cells. The R package, code, and analysis results used in this paper are available on https://github.com/doctorlaby/my-project. Conclusion: Our research is the first to utilize machine learning techniques in identifying DLST and LSP1 as significant biomarkers of VTE. With our findings, we have uncovered new insights into the underlying causes of VTE and potential treatments for affected patients.