Structure-Function Aware Evolutionary Multitasking for Therapeutic Peptide Co-discovery
提出一种结构功能感知的进化多任务算法SFEM，通过物理信息知识迁移策略，同时发现抗菌肽和抗癌肽，在18个协同发现任务上优于现有方法，并通过分子对接和动力学模拟验证了生物活性。
Therapeutic peptides have surfaced as promising new drugs for many diseases. Yet the discovery of novel therapeutic peptides suffers from the search of vast sequence space, the lack of validated data, and the complex physicochemical constraints. Co-discovering peptides of related therapeutic functions provides an opportunity to counteract the aforementioned limitations by leveraging knowledge transfer from well-studied peptides to underexplored ones. This paper introduces a Structure-Function aware Evolutionary Multitasking algorithm named SFEM based on a physics-informed knowledge transfer strategy for the simultaneous discovery of antimicrobial peptides (AMPs) and anticancer peptides (ACPs). Both structural and functional properties (i.e., membrane destructive effect and molecular activities) of the peptides are considered as the optimization objectives. SFEM is featured by an efficient anchor search utilizing amino acid distribution to update the residue of functionally analogous sequence regions across tasks, and a Markov process leveraging <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">n</i>-gram frequency distributions to capture and transfer contextual sub-sequence. SFEM was evaluated on a set of 18 AMP-ACP co-discovery tasks and demonstrated to achieve superior performance to other state-of-the-art methods. The bioactivities of the identified peptides were also validated through molecular docking and dynamics simulation, suggesting the effectiveness of SFEM.