Particle Swarm Optimization-Assisted Multilayer Ensemble Model to predict DNA 4mC sites

Abstract

DNA methylation is an epigenetic modification that plays a crucial role in various biological processes, including gene expression regulation, cell differentiation, and the development of diseases such as cancer. Identifying DNA methylation patterns is essential for understanding its functional implications. Traditional experimental methods for detecting DNA methylation are costly, time-consuming, and inefficient for analyzing large-scale sequencing data. In this research, we explore the application of machine learning techniques to accurately identify DNA methylation sites. Our research aims to develop a Particle Swarm Optimization-Assisted Multilayer Ensemble Model (PSO-MEM) with several significant contributions. These include extracting semantic features from genetic sequences, optimizing feature dimensions to reduce classification errors, developing a multilayer dynamic approach that transfers learned information between layers during classification, and incorporating ensemble techniques for improved prediction and optimal results. To evaluate the performance of our proposed model, we compare it with existing models using eight publicly available datasets. The results demonstrate the efficacy of our approach, with AUC values of 91.99%, 92.80%, 90.28%, 91.03%, 93.09%, 90.79%, 90.68%, and 91.88% for the C. elegans, D. melanogaster, A. thaliana, E. coli, G. subterraneus, G. pickeringi, F. vesca, and R. chinensis datasets, respectively. The results highlight the potential of machine learning techniques for efficient and reliable identification of DNA methylation sites in large-scale genomic data, facilitating advancements in understanding epigenetic modifications and their functional implications.