Non-Codeing RNAs Family Prediction Based on RNA Representation and Deep Learning

Predicting RNAs family presents a significant challenge with broad implications in medicine and scientific inquiry. Leveraging the advancements in deep learning, recent studies have delved into utilizing these algorithms for RNAs family prediction. This article introduces an innovative algorithm integrating BiLSTM, transformer, and convolutional neural networks (CNN). Initially, RNA sequences undergo representation using k-mers, a strategy aimed at mitigating the impact of errors in the sequence modeling process. Following this, a word embedding technique is applied to represent the RNA sequences, thereby reducing computational complexity within the network. Experimental results demonstrate the superior performance of our model compared to other comparison algorithms in terms of recall, accuracy, and precision on the 10-fold test dataset. This demonstrates the excellent comprehensive performance of the proposed model in terms of robustness and efficiency.