Actin Gamma 1, a new skin cancer pathogenic gene, identified by the biological feature-based classification.

Skin cancer is the most common form of cancer that accounting for at least 40% of cancer cases around the world. This study aimed to identify skin cancer-related biological features and predict skin cancer candidate genes by employing machine learning based on biological features of known skin cancer genes. The known skin cancer-related genes were fetched from database and encoded by the enrichment scores of gene ontology and pathways. The optimal features of the skin cancer related genes were selected with a series of feature selection methods, such as mRMR, IFS, and Random Forest algorithm. Quantitative PCR (Q-PCR) was performed for the predicted genes. Effects on proliferation and metastasis of skin cancer cell line A431 were detected through MTT and transwell assay. The effects on myosin light chain (MLC) phosphorylation of Actin Gamma 1 (ACTG1) were detected by Western blot. A total of 1233 GO terms and 55 KEGG pathway terms were identified as the optimal features for the depiction of skin cancer. According to those terms, 1134 possible skin cancer-related genes were predicted. We further identified 16 new biomarkers in expression and the classification model can predict skin cancer cases with 100% accuracy. Among the 16 genes, ACTG1 had significantly high expression in skin cancer tissue. Our investigation suggested that ACTG1 can regulate the cell proliferation and migration through ROCK signaling pathway.