RESUMEN
Human papillomavirus (HPV) infections often show no symptoms but sometimes lead to either warts or carcinoma based on the HPV genotype. The relationship between HPV infections and cervical cancer have been well studied in the past two decades. However, distinguishing carcinogenic HPV variants from non-carcinogenic ones remains a major challenge in clinical genetic testing of HPV-induced cancer samples. All of the published HPV carcinogenicity prediction methods are neither publically available nor tested with two-thirds of available HPV variants. The nucleotide composition-based studies are the simplest and most precise methods of characterizing new genomes. Hence, there is a need for machine learning models which can predict the carcinogenic nature of newly discovered HPV based on their genomic composition. We developed a standalone and web tool, CarcinoHPVPred (h t t p :// test5.bicpu.edu.in/CarcinoHPVPred.php), for predicting the phenotype of HPV with a range of a high accuracy between 94% - 100%. This tool consists of machine learning models build upon genomic features of two genes namely E2 and E6. Overall, the accurate and early prediction of carcinogenic nature of HPV can be performed with this only available tool of its kind till date.