Pattern Recognition of Core-Promoter DNA Sequences Based on Recurrent Neural Network

Abstract

The interpretation of genomic sequences remains a major challenge in computational biology due to the inability of traditional methods to detect complex, context-dependent regulatory patterns. therefore, detecting the promoter sequences tends to lead to a high false-positive rate. Additionally, to overcome these limitations, we used the Recurrent Neural Network (RNN) framework that autonomously identifies functional genomic elements by modelling long-range dependencies in Deoxyribonucleic Acid (DNA). we proposed the method of combining advanced pattern recognition with experimental validation, outperforming conventional techniques in detecting regulatory motifs while enabling standardized, high-throughput genomic annotation. By bridging computational and molecular biology, this approach provides a powerful solution, including synthetic biology and genome annotation pipelines. Benchmarking results demonstrate that the framework significantly improves detection of non-canonical and weakly conserved regulatory features, which are frequently missed by existing tools. To perform an analysis of the publicly available datasets: GRCh38.p14 on our proposed work, we have analyzed the accuracy is 92.26% in classifying genes and non-genes from the long DNA sequence.