Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
Review
. 2025 Jan;18(1):e70072.
doi: 10.1111/1751-7915.70072.

AI Methods for Antimicrobial Peptides: Progress and Challenges

Carlos A Brizuela  1 , Gary Liu  2 , Jonathan M Stokes  2 , Cesar de la Fuente-Nunez  3   4   5   6
Affiliations
Review

AI Methods for Antimicrobial Peptides: Progress and Challenges

Carlos A Brizuela et al. Microb Biotechnol. 2025 Jan.

Abstract

Antimicrobial peptides (AMPs) are promising candidates to combat multidrug-resistant pathogens. However, the high cost of extensive wet-lab screening has made AI methods for identifying and designing AMPs increasingly important, with machine learning (ML) techniques playing a crucial role. AI approaches have recently revolutionised this field by accelerating the discovery of new peptides with anti-infective activity, particularly in preclinical mouse models. Initially, classical ML approaches dominated the field, but recently there has been a shift towards deep learning (DL) models. Despite significant contributions, existing reviews have not thoroughly explored the potential of large language models (LLMs), graph neural networks (GNNs) and structure-guided AMP discovery and design. This review aims to fill that gap by providing a comprehensive overview of the latest advancements, challenges and opportunities in using AI methods, with a particular emphasis on LLMs, GNNs and structure-guided design. We discuss the limitations of current approaches and highlight the most relevant topics to address in the coming years for AMP discovery and design.

PubMed Disclaimer

Conflict of interest statement

Cesar de la Fuente‐Nunez provides consulting services to Invaio Sciences and is a member of the Scientific Advisory Boards of Nowture S.L., Peptidus and Phare Bio. De la Fuente is also on the Advisory Board of the Peptide Drug Hunting Consortium (PDHC). The de la Fuente Lab has received research funding or in‐kind donations from United Therapeutics, Strata Manufacturing PJSC and Procter & Gamble, none of which were used in support of this work. Other authors have no conflicts to declare. Jonathan Stokes is co‐founder and CSO of Stoked Bio.

Figures

FIGURE 1
FIGURE 1
Hierarchical diagram for AMP identification. An example input sequence is ‘GRACIAS’. The first level requires discrimination between AMPs and non‐AMPs. At the second level, the target is predicted to be active against viruses, bacteria and/or fungi. The third level asks for an additional layer of granularity, such as distinguishing between Gram‐positive and Gram‐negative bacteria. The fourth level further discriminates between species. Finally, the fifth level provides detailed information on the minimum inhibitory concentration (MIC).
FIGURE 2
FIGURE 2
Graph neural network for AMPS. (a) Input peptide MLGYF, PDB Id 2NOU. (b) Node (x→) and edge (y→) descriptors of variable size. (c) A graph representation for the peptide MLGYF.
FIGURE 3
FIGURE 3
(a) Number of publications per year of works on LLMs. Query = ‘Large Language Models’. (b) Number of publications per year of works on Protein LMs. Query = ‘Protein Language Models’.
FIGURE 4
FIGURE 4
General framework for the use of LLMs in the peptide context. Stage 1 consists of a tokenizer, embedding layer and the transformer layer (this layer is absent in some approaches). Stage 2 receives the output of the embedding layer or of the transformer layer, then by using this representation, the fine‐tuning layer is trained with for the specific learning task such as antimicrobial activity discrimination.
FIGURE 5
FIGURE 5
General scheme for evolutionary optimisation‐based approaches. The fitness function can be related to physicochemical properties the design wants to achieve (phenomenological). The fitness can be guided by an ML repressor (ML‐based), or it can use some target information (Target similarity). WLE stands for Wet Lab Experiments and MDS for molecular dynamics simulation. The fitness function can also utilise any combination of the three approaches.
See this image and copyright information in PMC

References

    1. Ageitos, J. M. , Sánchez‐Pérez A., Calo‐Mata P., and Villa T. G.. 2017. “Antimicrobial Peptides (AMPs): Ancient Compounds That Represent Novel Weapons in the Fight Against Bacteria.” Biochemical Pharmacology 133: 117–138. - PubMed
    1. Aguilera‐Mendoza, L. , Marrero‐Ponce Y., Beltran J. A., Tellez Ibarra R., Guillen‐Ramirez H. A., and Brizuela C. A.. 2019. “Graph‐Based Data Integration From Bioactive Peptide Databases of Pharmaceutical Interest: Toward an Organized Collection Enabling Visual Network Analysis.” Bioinformatics 35: 4739–4747. - PubMed
    1. Aguilera‐Puga, M. D. C. , Cancelarich N. L., Marani M. M., De la Fuente‐Nunez C., and Plisson F.. 2024. “Accelerating the Discovery and Design of Antimicrobial Peptides With Artificial Intelligence.” In Computational Drug Discovery and Design, edited by Gore M. and Jagtap U. B., 329–352. New York, NY: Springer US. - PMC - PubMed
    1. Baek, M. , DiMaio F., Anishchenko I., et al. 2021. “Accurate Prediction of Protein Structures and Interactions Using a Three‐Track Neural Network.” Science 373, no. 6557: 871–876. - PMC - PubMed
    1. Bahar, A. A. , and Ren D.. 2013. “Antimicrobial Peptides.” Pharmaceuticals 6, no. 12: 1543–1575. - PMC - PubMed

Publication types

Substances

LinkOut - more resources