Can structural analysis identify new toxins? This article introduces a graphical model to identify pore-forming toxins (PFTs), proteins that create lesions in biological membranes. Overcoming limitations of sequence homology-based methods, the approach constructs a protein structure graph based on consensus secondary structures and develops a semi-Markov conditional random fields model for protein sequence segmentation. The method distinguishes structurally similar proteins even with low sequence identity, a feat unattainable by traditional approaches. Additionally, an efficient framework for UniRef50 aids in extracting proteins of interest for further study.
This paper is well-suited for Proteins: Structure, Function, and Bioinformatics. By proposing a graphical model for identifying pore-forming proteins based on structural analysis, it directly addresses protein structure and function, a core area of the journal's focus.