{"title":"Discovery of antimicrobial peptides from Bacillus genomes against phytopathogens with deep learning models","authors":"Huan Su, Mengli Gu, Zechao Qu, Qiao Wang, Jingjing Jin, Peng Lu, Jianfeng Zhang, Peijian Cao, Xueliang Ren, Jiemeng Tao, Boyang Li","doi":"10.1186/s40538-025-00751-9","DOIUrl":null,"url":null,"abstract":"<div><h3>Background</h3><p>Antimicrobial peptides (AMPs) have emerged as a potential novel class of antimicrobial agents due to their broad microbial targeting and low resistance risks. Although AMPs have limited applications in agriculture, their potential to replace chemical pesticides could address food security and environmental concerns. Members of <i>Bacillus</i> sp., abundant in soil and plant microbiomes, are recognized as important sources of AMPs for their resistance and strong antimicrobial properties, making them ideal candidates for biocontrol in sustainable agriculture. To harness this potential, this study employed deep learning models to predict AMPs derived from <i>Bacillus</i> genomes, aiming to identify candidates with high activity against phytopathogens. Subsequently, the antimicrobial efficacy of selected candidate AMPs was experimentally validated.</p><h3>Results</h3><p>More than 6700 <i>Bacillus</i> genomes were collected to identify a broad range of short peptides (10–100 amino acids), which were analyzed using advanced deep learning models, including BERT, Mamba, CNN-LSTM, and CNN-Attention. These models demonstrated enhanced predictive accuracy and reliability over existing methods, and resulted in 4,993,389 potential AMPs from <i>Bacillus</i> genomes. Among these AMPs, two high-confidence AMPs (cAMP_1 and cAMP_2) were selected by cross-validation, and their structural stability and activity were evaluated and verified by molecular dynamics simulations and experimental assays, respectively. Both of them exhibited antimicrobial activity against <i>Escherichia coli</i>, <i>Staphylococcus aureus</i>, and various common agricultural fungal and bacterial pathogens.</p><h3>Conclusions</h3><p>This high-throughput deep learning pipeline successfully uncovered novel AMPs from <i>Bacillus</i> genomes, which underscored the efficiency of deep learning models in identifying functional peptides. This approach could accelerate the discovery of potential AMPs for biocontrol applications in plant disease management, contributing to sustainable agriculture and reduced dependency on traditional antibiotics.</p><h3>Graphical abstract</h3>\n<div><figure><div><div><picture><source><img></source></picture></div></div></figure></div></div>","PeriodicalId":512,"journal":{"name":"Chemical and Biological Technologies in Agriculture","volume":"12 1","pages":""},"PeriodicalIF":5.2000,"publicationDate":"2025-03-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://chembioagro.springeropen.com/counter/pdf/10.1186/s40538-025-00751-9","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Chemical and Biological Technologies in Agriculture","FirstCategoryId":"97","ListUrlMain":"https://link.springer.com/article/10.1186/s40538-025-00751-9","RegionNum":2,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"AGRICULTURE, MULTIDISCIPLINARY","Score":null,"Total":0}
引用次数: 0
Abstract
Background
Antimicrobial peptides (AMPs) have emerged as a potential novel class of antimicrobial agents due to their broad microbial targeting and low resistance risks. Although AMPs have limited applications in agriculture, their potential to replace chemical pesticides could address food security and environmental concerns. Members of Bacillus sp., abundant in soil and plant microbiomes, are recognized as important sources of AMPs for their resistance and strong antimicrobial properties, making them ideal candidates for biocontrol in sustainable agriculture. To harness this potential, this study employed deep learning models to predict AMPs derived from Bacillus genomes, aiming to identify candidates with high activity against phytopathogens. Subsequently, the antimicrobial efficacy of selected candidate AMPs was experimentally validated.
Results
More than 6700 Bacillus genomes were collected to identify a broad range of short peptides (10–100 amino acids), which were analyzed using advanced deep learning models, including BERT, Mamba, CNN-LSTM, and CNN-Attention. These models demonstrated enhanced predictive accuracy and reliability over existing methods, and resulted in 4,993,389 potential AMPs from Bacillus genomes. Among these AMPs, two high-confidence AMPs (cAMP_1 and cAMP_2) were selected by cross-validation, and their structural stability and activity were evaluated and verified by molecular dynamics simulations and experimental assays, respectively. Both of them exhibited antimicrobial activity against Escherichia coli, Staphylococcus aureus, and various common agricultural fungal and bacterial pathogens.
Conclusions
This high-throughput deep learning pipeline successfully uncovered novel AMPs from Bacillus genomes, which underscored the efficiency of deep learning models in identifying functional peptides. This approach could accelerate the discovery of potential AMPs for biocontrol applications in plant disease management, contributing to sustainable agriculture and reduced dependency on traditional antibiotics.
期刊介绍：
Chemical and Biological Technologies in Agriculture is an international, interdisciplinary, peer-reviewed forum for the advancement and application to all fields of agriculture of modern chemical, biochemical and molecular technologies. The scope of this journal includes chemical and biochemical processes aimed to increase sustainable agricultural and food production, the evaluation of quality and origin of raw primary products and their transformation into foods and chemicals, as well as environmental monitoring and remediation. Of special interest are the effects of chemical and biochemical technologies, also at the nano and supramolecular scale, on the relationships between soil, plants, microorganisms and their environment, with the help of modern bioinformatics. Another special focus is the use of modern bioorganic and biological chemistry to develop new technologies for plant nutrition and bio-stimulation, advancement of biorefineries from biomasses, safe and traceable food products, carbon storage in soil and plants and restoration of contaminated soils to agriculture.
This journal presents the first opportunity to bring together researchers from a wide number of disciplines within the agricultural chemical and biological sciences, from both industry and academia. The principle aim of Chemical and Biological Technologies in Agriculture is to allow the exchange of the most advanced chemical and biochemical knowledge to develop technologies which address one of the most pressing challenges of our times - sustaining a growing world population.
Chemical and Biological Technologies in Agriculture publishes original research articles, short letters and invited reviews. Articles from scientists in industry, academia as well as private research institutes, non-governmental and environmental organizations are encouraged.
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号
