Fig. 3: Key AA features and key feature fragments. | Nature Communications

Fig. 3: Key AA features and key feature fragments.

From: DLFea4AMPGen de novo design of antimicrobial peptides by integrating features learned from deep learning models

Fig. 3

a The length distribution of 4760 samples with potential antibacterial, antifungal, and antioxidant activities, with sequences of 13 AAs appearing most frequently. b Illustration of KFFs extraction. A window of 13 AAs is set to slide across the entire sequence with a step size of 1 AA. Fragments with the highest sum of average SHAP values and scores greater than 0 are selected. 3400 fragments that meet all conditions are being focused on as KFFs. c Comparison of the proportion of each AA in KFFs and in all positive samples across 20 bioactive peptide datasets. d Distribution of 3400 KFFs in four subfamilies observed across the phylogenetic tree, which were subfamily-A (SF-A), subfamily-B (SF-B), subfamily-C (SF-C), and subfamily-D (SF-D). The colored dots at the tips of the branches indicate the original bioactive peptides source of each KFF. e AA distribution at the 13 positions in each KFF subfamily, and the highlighted AAs at each position are the top three AAs that most frequently occur.

Back to article page