Scientists used a big peptide database to design short proteins that can kill HIV or antibiotic‑resistant bacteria. They found anti‑HIV peptides need many arginine residues and a disulfide link, while anti‑MRSA peptides are straight chains rich in leucine. The detailed 3‑D structure of the natural peptide LL‑37 was solved, giving clues on how it attacks microbes.

Antimicrobial peptides (AMPs), small host defense proteins, are indispensable for the protection of multicellular organisms such as plants and animals from infection. The number of AMPs discovered per year increased steadily since the 1980s. Over 2,000 natural AMPs from bacteria, protozoa, fungi, plants, and animals have been registered into the antimicrobial peptide database (APD). The majority of these AMPs (>86%) possess 11-50 amino acids with a net charge from 0 to +7 and hydrophobic percentages between 31-70%. This article summarizes peptide discovery on the basis of the APD. The major methods are the linguistic model, database screening, de novo design, and template-based design. Using these methods, we identified various potent peptides against human immunodeficiency virus type 1 (HIV-1) or methicillin-resistant Staphylococcus aureus (MRSA). While the stepwise designed anti-HIV peptide is disulfide-linked and rich in arginines, the ab initio designed anti-MRSA peptide is linear and rich in leucines. Thus, there are different requirements for antiviral and antibacterial peptides, which could kill pathogens via different molecular targets. The biased amino acid composition in the database-designed peptides, or natural peptides such as θ-defensins, requires the use of the improved two-dimensional NMR method for structural determination to avoid the publication of misleading structure and dynamics. In the case of human cathelicidin LL-37, structural determination requires 3D NMR techniques. The high-quality structure of LL-37 provides a solid basis for understanding its interactions with membranes of bacteria and other pathogens. In conclusion, the APD database is a comprehensive platform for storing, classifying, searching, predicting, and designing potent peptides against pathogenic bacteria, viruses, fungi, parasites, and cancer cells.