Machine Learning applied to an RNA Classification Problem

Authors

DOI:

https://doi.org/10.55632/pwvas.v94i1.895

Abstract

Long RNA sequences can be classified as protein-coding messenger RNA (mRNA) or as long non-coding RNA (lncRNA). RNA classification based on sequence alone is a bioinformatics challenge with potential to enlarge our understanding of the causes of many human diseases. We chose to compare two classification approaches: (1) applications of correlation statistics and visualization tools, and (2) training a convolutional neural network. Both approaches used human RNA sequences from the public GenCode database. We found that the machine learning approach was superior, achieving an accuracy of 87.37%, which is not as high as some published classifiers. These results indicate that machine learning techniques are a more effective solution to the problem of human RNA classification compared to the other techniques tested. For future work, we propose to build, train, and compare other convolutional neural network models for this classification task.

Author Biographies

Andrew Moseman, Shepherd University

Computer science student

Samon Bhatta, Shepherd University

Computer science student

Jason Rafe Miller, Shepherd University

Assistant Professor of Computer Science

Department of Computer Science, Mathematics, and Engineering

Downloads

Published

2022-04-22

How to Cite

Moseman, A., Bhatta, S., & Miller, J. R. (2022). Machine Learning applied to an RNA Classification Problem. Proceedings of the West Virginia Academy of Science, 94(1). https://doi.org/10.55632/pwvas.v94i1.895

Issue

Section

Meeting Abstracts-Poster