Identification of efficient learning classifiers for discrimination of coding and non-coding RNAs in plant species

Abstract

Though the non-coding RNAs (ncRNAs) do not encode for proteins, they act as functional RNAs and regulate gene expression besides their involvement in disease-causing mechanisms and epigenetic mechanisms. Thus, discriminating ncRNAs from coding RNAs (cRNAs) is important in transcriptome studies. Several machine learning-based classifiers, including deep learning classifiers, have been employed for discriminating cRNAsfrom ncRNAs. However, the performance comparison of such classifiers in plant species is yet to be ascertained. Thus, in the present study, the performance of the classifiers such as Deep Neural Network (DNN), Random Forest (RF), Support Vector Machine (SVM), and Artificial Neural Network (ANN) were evaluated for classifying cRNAs and ncRNAsby using the datasets of plant species including crops such as rice, wheat, maize, cotton, sunflower, barley, banana, grape, papaya. Further, the performance of classifiers was assessed by following the cross-validation process as well as by considering an independent test data set of 3,997 cRNAs and 4,110 ncRNAs. The results revealed that Random Forest classifier exhibited highest performance accuracy (99.803%) among the machine learning classifiers, followed by DNN (99.519%), SVM (97.364%) and ANN (99.260%). The present study is expected to help computational and experimental biologists for easy discrimination between coding and non-coding RNAs.