Design of selective peptide antibiotics by using the sequence moment concept
New antibiotics against multidrug-resistant bacteria are urgently needed, but rapid acquisition of resistance limits their usefulness. Endogenous antimicrobial peptides (AMPs) with moderate selectivity, but multimodal mechanism of action, have remained effective against bacteria for millions of years. Their therapeutic application, however, requires optimizing the balance between antibacterial activity and selectivity, so that rational design methods for increasing selectivity are highly desirable. We have created training (n=36) and testing (n=37) sets from frog-derived AMPs with determined therapeutic index (TI). The 'sequence moments' concept then enabled us to find a one-parameter linear model resulting in a good correlation between measured and predicted TI (r2=0.83 and 0.64 for each set, respectively). The concept was then used in the AMP-Designer algorithm to propose primary structures for highly selective AMPs against Gram-negative bacteria. Testing the activity of one such peptide produced a TI>200 as compared to the best AMP in the data-base, with TI=125.