Deep-m6Am: a deep learning model for identifying N6, 2′-O-Dimethyladenosine (m6Am) sites using hybrid features

Abstract

N6,2′-O-dimethyladenosine (m6Am) is a crucial RNA modification that plays a pivotal role in regulating gene expression and maintaining RNA stability. Given its dynamic involvement in various biological processes and diseases, accurately identifying m6Am is essential for understanding cellular mechanisms and pathogenesis. Furthermore, detecting m6Am modifications is key to deciphering regulatory pathways and elucidating disease mechanisms. In this study, we propose Deep-m6Am, a deep learning–based model for precisely identifying m6Am sites in RNA sequences. The proposed framework employs a comprehensive feature extraction process, i.e., integrating pseudo single nucleotide composition (PseSNC), pseudo dinucleotide composition (PseDNC), and pseudo trinucleotide composition (PseTNC) to capture complex sequence patterns. To enhance computational efficiency and eliminate noisy or redundant features, a supervised SHAP (SHapley Additive exPlanations) algorithm is utilized, ensuring the selection of the most informative features. Finally, a multilayer deep neural network (DNN) is used as a classification algorithm for identifying m6Am sites. The performance of the proposed model was evaluated in comparison with traditional machine learning (ML) algorithms and existing models. Experimental results demonstrate that Deep-m6Am outperforms previous approaches by 6.67% and traditional ML algorithms by 7.39%.These findings highlight Deep-m6Am as a promising tool for advancing drug discovery and improving the diagnosis of diseases associated with m6Am modifications.

Keywords

• machine learning
• deep neural network
• hybrid sequential model
• deep-m6am model
• n6, 2′o-dimethyl adenosine
• rna modification