Abstract
BACKGROUND: Lysine post-translational modification (PTM) is pivotal in regulating diverse cellular processes, profoundly impacting protein structure and function. Over recent decades, numerous experimental techniques have advanced PTM site identification, significantly contributing to research progress. However, these methods are time-intensive and labor-intensive. Deep learning technologies have shown promise in predicting PTM sites, yet current approaches struggle with species-specific PTM site prediction. METHODS: We introduce MDDeep-Ace, a novel deep learning method based on multi-domain adaptation for predicting lysine acetylation sites. By integrating data from multiple species, MDDeep-Ace enhances the generalization of species-specific prediction models, improving predictive performance. RESULTS: Experimental findings illustrate that our proposed multi-domain adaptation approach significantly enhances prediction accuracy across multiple species, surpassing existing lysine acetylation site prediction tools.