EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Surrogate-guided self-play Monte Carlo tree search for antimicrobial peptide sequence optimization and in vitro antibacterial assessment

Jang-Jih Lu, Chia-Ru Chung iD, Chen-Hao Chang, Li-Ching Wu iD, Justin Bo-Kai Hsu, Cheng-Yu Tsai, Tzong-Yi Lee iD, Jorng-Tzong Horng

Briefings in Bioinformatics · 02.11.2026 · 27 (6)

DOI: 10.1093/bib/bbag618 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Briefings in Bioinformatics, online erschienen am 13.11.2026. Verfasst von Jang-Jih Lu und 7 weiteren Autor:innen. Beteiligte Einrichtung: Division of Clinical Pathology, Taipei Tzu Chi Hospital, Buddhist Tzu Chi Medical Foundation, No. 289, Jianguo Rd., Xindian District, New Taipei City…. Hauptthemen laut MEDLINE: Antimicrobial Peptides, Anti-Bacterial Agents. Die Arbeit nennt 41 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Project for an Intelligent Multi-Omics Precision Medicine, Ministry of Science and Technology, Taiwan, Taipei Tzu Chi Hospital, Buddhist Tzu Chi Medical Foundation.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The rapid emergence of multidrug-resistant pathogens necessitates the development of novel therapeutic agents, yet the de novo design of antimicrobial peptides (AMPs) is challenged by the vast combinatorial search space and the difficulty of balancing efficacy with toxicity. To address these limitations, we introduce a reinforcement learning framework that employs surrogate-guided self-play Monte Carlo tree search to explore the peptide landscape. By integrating surrogate models for predicted antimicrobial activity, HemoDL-predicted hemolytic propensity, and ESMFold structural confidence, the system performs multi-property computational optimization and candidate prioritization. In silico analysis demonstrated that the optimized peptides adopt amphipathic α-helical structures with high predicted folding confidence and possess favorable docking scores against Staphylococcus aureus resistance targets, including PBP2a and MurA. To assess the biological relevance of the computationally prioritized candidates, 16 peptides were synthesized and evaluated by agar dilution against a 37-strain bacterial panel over a concentration range of 0.25–64 μg/ml. AMP2, AMP7, and AMP9 achieved complete inhibition of at least one tested strain, corresponding to 3 of 16 synthesized candidates (18.75%). AMP12 and AMP13 showed partial growth reduction without complete inhibition within the tested range, whereas the remaining 11 candidates showed no complete inhibition. These findings quantify the current gap between surrogate-defined computational enrichment and assay-specific antibacterial activity. The strain-dependent experimental outcomes further indicate that surrogate, docking, and structural scores do not fully capture all determinants of antibacterial efficacy. The experimental assessment was limited to bacterial agar-dilution activity; mammalian-cell cytotoxicity, wet-lab hemolysis, and peptide-stability assays were not performed. In conclusion, this study establishes a computational framework for the design of bioactive peptides and provides structural insights for bridging the gap between deep learning predictions and experimental outcomes.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Antimicrobial PeptidesAnti-Bacterial AgentsMonte Carlo MethodStaphylococcus aureusMicrobial Sensitivity TestsAmino Acid SequenceReinforcement Machine LearningHumansMolecular Docking SimulationHemolysis

Stichwörter der Autor:innen

AlphaZero-inspired designdeep learning surrogatesminimum inhibitory concentrationmulti-property optimizationreinforcement learning

Substanzen

Antimicrobial PeptidesAnti-Bacterial Agents

Autor:innen und Institutionen

  • Division of Clinical Pathology, Taipei Tzu Chi Hospital, Buddhist Tzu Chi Medical Foundation, No. 289, Jianguo Rd., Xindian District, New Taipei City 23142, Taiwan.
  • Department of Computer Science and Information Engineering, National Central University, No. 300, Zhongda Rd., Zhongli District, Taoyuan City 320317, Taiwan.
  • Department of Biomedical Sciences and Engineering, National Central University, No. 300, Zhongda Rd., Zhongli District, Taoyuan City 320317, Taiwan.
  • Department of Computer Science and Engineering, Yuan Ze University, No. 135, Yuan-Tung Road, Zhongli District, Taoyuan City 320315, Taiwan.
  • Institute of Bioinformatics and Systems Biology, National Yang Ming Chiao Tung University, No. 75, Boai Street, East District, Hsinchu City 300, Taiwan.
  • Center for Intelligent Drug Systems and Smart Biodevices (IDS2B), National Yang Ming Chiao Tung University, No. 75, Boai Street, East District, Hsinchu City 300, Taiwan.

ORCID-Profile: Chia-Ru Chung, Li-Ching Wu, Tzong-Yi Lee

Förderung

  • Project for an Intelligent Multi-Omics Precision Medicine
  • Ministry of Science and Technology, Taiwan
  • Taipei Tzu Chi Hospital, Buddhist Tzu Chi Medical Foundation
  • National Science and Technology Programs

Zitieren

APA

Lu, J. J., Chung, C. R., Chang, C. H., Wu, L. C., Hsu, J. B., Tsai, C. Y., Lee, T. Y., & Horng, J. T. (2026). Surrogate-guided self-play Monte Carlo tree search for antimicrobial peptide sequence optimization and in vitro antibacterial assessment. Briefings in Bioinformatics, 27(6). https://doi.org/10.1093/bib/bbag618

Vancouver

Lu JJ, Chung CR, Chang CH, Wu LC, Hsu JB, Tsai CY, et al. Surrogate-guided self-play Monte Carlo tree search for antimicrobial peptide sequence optimization and in vitro antibacterial assessment. Briefings in Bioinformatics. 2026;27(6). doi:10.1093/bib/bbag618

BibTeX
@article{Lu2026,
  author = {Lu, Jang-Jih and Chung, Chia-Ru and Chang, Chen-Hao and Wu, Li-Ching and Hsu, Justin Bo-Kai and Tsai, Cheng-Yu and Lee, Tzong-Yi and Horng, Jorng-Tzong},
  title = {Surrogate-guided self-play Monte Carlo tree search for antimicrobial peptide sequence optimization and in vitro antibacterial assessment},
  journal = {Briefings in Bioinformatics},
  year = {2026},
  volume = {27},
  number = {6},
  doi = {10.1093/bib/bbag618}
}
Kostenlos starten →