Detecting N6-methyladenosine sites from RNA transcriptomes using random forest

Abstract

N6-methyladenosine (m6A) modifications are one the most frequently occurring RNA post transcriptional modifications. These modifications perform vital roles in different biological processes, including, localization and translation of proteins, X chromosome inactivation, cell stability, microRNA regulation, and reprogramming etc. Any abnormal change in m6A sites may lead to several abnormalities, including, cancer, brain-related disorders and many other life threatening diseases. Precise detection of m6A modifications is crucial for the diagnosis and treatment of these diseases. Existing methods suffer from the problem of inefficient detection of m6A sites, especially in yeast transcriptomes (due to varied structure) and inability of the computational techniques to capture the encoded information surrounding the m6A sites. In this work, we propose a novel method (called m6A-pred predictor) that utilizes a fusion of characteristics including, statistical, and chemical properties of the nucleotides, to precisely predict the presence of m6A sites in RNA sequences. The fusion of multiple types of features results in a high dimensional vector which is further optimized using an evolutionary algorithm. Finally, the random forest classifier is used to detect m6A sites by using the most discriminative features. The results, benchmarked on yeast transcriptomes, indicate that m6A-pred predictor outperforms all the previously reported predictors, notably, with an accuracy value of 78.58%, specificity value of 79.65% and Matthews correlation coefficient of 0.5717.