Abstract
Proto-oncogenes are the genes that have the potential to transform normal cells into cancer cells as a result of mutations. They usually contain encoding of proteins whose function is to inhibit cell differentiation, stimulate cell division, and prevent the death of cells. While the prognosis regarding proto-oncogene may occur at varying phases of cancer, the accuracy of the identification method is always questionable. The standard procedure for detecting these genes involves in-vitro experimentations but it proves to be very costly, time taking, and laborious. This problem is addressed by the use of computer-aided approaches established in studies encompassing methods in computational biology and bioinformatics. Early diagnosis of cancer is crucial for the full recovery of the patient. Proto-oncogene proteins are an important biomarker that helps identify the onset of a specific type of cancer. Keeping this in mind, this study proposes an efficient methodology for in-silico identification of proto-oncogenes. The predictor proposed in this study computes position and composition relevant statistical features incorporated into the pseudo-amino-acid composition (PseAAC) based on Chou's 5-step rules. Subsequently, the study finds that the use of a random forest classifier performs the most accurate prediction of proto-oncogene proteins. The method was validated using the 10 folds cross-validation, Jackknife testing, Self-Consistency, and Independent set testing, giving 95.44%, 97.17%, 99.8%, and 96.41% accurate results, respectively. These results imply that the proposed model can play a key role in the early prognosis of cancer and aid scientists in the discovery of mechanisms against cancer.