Scholarlink

Purpose <p>The global health crisis of antibiotic-resistant infections, particularly in chronic wounds, demands innovative therapeutic strategies. Antimicrobial peptides (AMPs) represent promising candidates due to their low resistance potential and multi-target mechanisms.</p> Methods <p>We developed an interpretable machine learning-guided QSAR model using a random forest (RF) algorithm based on 2D molecular descriptors to predict the minimum inhibitory concentration (MIC) of AMPs against <i>Staphylococcus aureus</i>. The dataset was curated from the DBAASP database and filtered based on MIC thresholds, peptide length (5–15 amino acids), and culture conditions. External validation yielded (R²<sub>test</sub> = 0.84, Q²F1 = 0.61, MAE = 0.18).</p> Results <p>Feature importance analysis revealed three key descriptors: maxHBint2 (hydrogen bonding capacity), SpMin7_Bhe, and SpMin6_Bhp (Burden eigenvalues reflecting molecular topology). Using this model, we designed and synthesized two novel pentapeptides, R2W3 and R2W2L, which exhibited potent anti-staphylococcal activity (MIC = 41.68 and 24.54 µM, respectively). In vitro assays confirmed low cytotoxicity (&lt; 10%) and &lt; 5% hemolysis. Molecular docking revealed distinct target specificities: R2W3 showed strong binding to <i>dihydrofolate reductase type 1</i> (ΔG = -16.84&#xa0;kcal/mol), while R2W2L preferentially interacted with <i>DNA gyrase subunit B</i> (ΔG = -8.39&#xa0;kcal/mol).</p> Conclusions <p>Our study establishes a robust computational-experimental pipeline for designing anti-<i>Staphylococcal</i> peptides with validated safety and efficacy. This approach provides a foundation for integrating multi-endpoint prediction and next-generation AMP discovery strategies.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Rational Design of Novel Anti-Staphylococcal Peptides Using Machine Learning-Guided QSAR Modeling and Experimental Validation

  • 期刊论文

  • 发表日期：

    2025年7月23日

查看全文
  • Saeed Zanganeh,
  • Alireza Farsinejad,
  • Ali Afgar,
  • Nasir Mohajel,
  • Mohamad Javad Mirzaei-Parsa,
  • Arman Shahabi

摘要

Purpose

The global health crisis of antibiotic-resistant infections, particularly in chronic wounds, demands innovative therapeutic strategies. Antimicrobial peptides (AMPs) represent promising candidates due to their low resistance potential and multi-target mechanisms.

Methods

We developed an interpretable machine learning-guided QSAR model using a random forest (RF) algorithm based on 2D molecular descriptors to predict the minimum inhibitory concentration (MIC) of AMPs against Staphylococcus aureus. The dataset was curated from the DBAASP database and filtered based on MIC thresholds, peptide length (5–15 amino acids), and culture conditions. External validation yielded (R²test = 0.84, Q²F1 = 0.61, MAE = 0.18).

Results

Feature importance analysis revealed three key descriptors: maxHBint2 (hydrogen bonding capacity), SpMin7_Bhe, and SpMin6_Bhp (Burden eigenvalues reflecting molecular topology). Using this model, we designed and synthesized two novel pentapeptides, R2W3 and R2W2L, which exhibited potent anti-staphylococcal activity (MIC = 41.68 and 24.54 µM, respectively). In vitro assays confirmed low cytotoxicity (< 10%) and < 5% hemolysis. Molecular docking revealed distinct target specificities: R2W3 showed strong binding to dihydrofolate reductase type 1 (ΔG = -16.84 kcal/mol), while R2W2L preferentially interacted with DNA gyrase subunit B (ΔG = -8.39 kcal/mol).

Conclusions

Our study establishes a robust computational-experimental pipeline for designing anti-Staphylococcal peptides with validated safety and efficacy. This approach provides a foundation for integrating multi-endpoint prediction and next-generation AMP discovery strategies.