What are antimicrobial peptides and how can AI help create them?
Antimicrobial peptides (AMPs) are small proteins produced by humans and many animal species to protect organisms from pathogens. Scientists have been studying AMPs for over a century because bacteria are very slow to develop resistance to them. However, rarely do AMPs prove to be more effective than antibiotics in the clinics. Recently, artificial intelligence models have been used in this research. A team led by Prof. Ewa Szczurek at the Faculty of Mathematics, Informatics, and Mechanics at the University of Warsaw has developed the HydrAMP algorithm*. Using a large set of amino acid chains, HydrAMP learns which sequences encode an active peptide desirable for fighting bacteria and which are inactive.
“Our algorithm can generate new antibacterial peptides. It also has the ability to ‘improve’ existing peptides so that they gain antibacterial activity through changes in the amino acid sequence,” explains Prof. Ewa Szczurek.
In this way, artificial intelligence can generate high-performance peptide sequences and suggest modifications to the sequences of less active peptides.
New peptides and their application confirmed in the laboratory
A great advantage of some AMPs is that they can combat truly dangerous and drug-resistant bacteria. Each newly discovered peptide offers hope for effectively combating bacteria such as Staphylococcus aureus. The activity of the peptides proposed by HydrAMP was first verified through molecular dynamics simulations, in collaboration with the team of Prof. Piotr Setny from the Centre of New Technologies at the University of Warsaw (CeNT UW). However, theoretical and computer simulations must be verified in the laboratory. The peptides were produced and tested at the Medical University of Gdańsk laboratory. The experiments carried out by Prof. Wojciech Kamysz’s team confirmed that HydrAMP transformed a completely inactive peptide into one with a very promising activity profile, without making it toxic. This is the first model of its kind capable of changing the peptide sequence, leading to increased antibacterial activity.
“Tests in Prof. Kamysz’s laboratory demonstrated the effect of some of our peptides on strains of Staphylococcus aureus, two strains of E. coli, as well as some of the most common sources of healthcare-associated infections: Acinetobacter baumannii and Pseudomonas aeruginosa,” adds Prof. Szczurek.
With the help of the HydrAMP algorithm, researchers have discovered a total of fifteen new peptides with high antibacterial activity, including one called Varsavian.
The text was originally published in Polish on the Serwis Naukowy UW website on February 4, 2025.
