Abstract
This study identified common genes associated with various primary cancers, including prostate cancer, and established a predictive model to accurately forecast the likelihood of cancer occurrence and its specific type. The aim is to offer physicians assistance in treatment decisions and meticulous follow-up, ultimately improving patient prognosis. Tumor sample data were collected from TCGA database, and differential expression analysis was employed for feature gene selection. Machine learning models were constructed to trace the origin of cancer genes. The results revealed 663 differentially expressed genes exhibiting characteristic expression in prostate cancer, squamous cell lung cancer, thyroid cancer, clear cell renal cell carcinoma, and bladder urothelial carcinoma. Logistic regression demonstrated superior stability and performance, with an average accuracy increase of 4% compared to other models. Therefore, precise prediction of cancer occurrence and its specific type based on gene expression status can be achieved, providing robust support for physicians' diagnosis and treatment decisions. This approach has the potential to enhance patient prognosis by enabling accurate predictions and targeted interventions.