Predicting Melanoma Staging using Targeted RNA Sequencing data using Machine Learning

Abstract

ABSTRACTThe aim of this study is to use machine learning to predict tumor staging and metastasis in melanoma with differentially expressed genes. Machine has been used in different clinical setting to predict different outcomes. However, it has not been used to look at predicting the diagnostic aspect of tumor staging. We used the TCGA RNA-Sequencing data on melanomas to predict tumor staging nodal and/or metastasis using deep neural networks (DNN) and random forest classifier (RF). Results: We were able to predict tumor staging (lower vs higher stage, i.e. Tis / T1 / T2 vs T3 and higher), nodal metastasis and combined nodal or distant metastasis in patients with melanomas with high accuracies. However, we need to further validate these results.