NCC—An Efficient Deep Learning Architecture for Non-Coding RNA Classification

Abstract

In this paper, an efficient deep-learning architecture is proposed, aiming to classify a significant category of RNA, the non-coding RNAs (ncRNAs). These RNAs participate in various biological processes and play an important role in gene regulation as well. Because of their diverse nature, the task of classifying them is a hard one in the bioinformatics domain. Existing classification methods often rely on secondary or tertiary RNA structures, which are computationally expensive to predict and prone to errors, especially for complex or novel ncRNA sequences. To address these limitations, a deep neural network classifier called NCC is proposed, which focuses solely on primary RNA sequence information. This deep neural network is appropriately trained to identify patterns in ncRNAs, leveraging well-known datasets, which are publicly available. Additionally, a ten times larger dataset than the available ones is created for better training and testing. In terms of performance, the suggested model showcases a 6% enhancement in precision compared to prior state-of-the-art systems, with an accuracy level of 92.69%, in the existing dataset. In the larger one, its accuracy rate exceeded 98%, outperforming all related tools, pointing to high prediction capability, which can act as a base for further findings in ncRNA analysis and the genomics field in general.