Scholarlink

Purpose <p>Quick identification of new molecules against protein involved in cancer&#xa0;pathogenesis by machine learning (ML) approach is the main goal of this research. The&#xa0;fibroblast growth factor receptor 1 (FGFR1) is a robust target in cancer therapy,&#xa0;with its&#xa0;overexpression linked to various malignancies. Despite the availability&#xa0;of FDA approved&#xa0;FGFR1 inhibitors, drug resistance and specificity issues highlight the need&#xa0;for novel therapeutic candidates.</p> Methods <p>This study integrates&#xa0;ML&#xa0;with molecular docking and&#xa0;molecular dynamics simulations (MDS) to identify and validate potential FGFR1&#xa0;inhibitors.</p> Results <p>ML models were trained and optimized to build a voting classifier using 1,523&#xa0;data points of FGFR1 known inhibitors, attaining an accuracy of 89% and an AUC&#xa0;score (area under the curve) of 0.95. The optimized model was applied to&#xa0;COCONUT database (695,133 molecules), yielding 12 potential inhibitors, among&#xa0;which ligand M6 exhibited the highest predicted activity. Molecular docking confirmed&#xa0;its stronger binding affinity (-9.55 kcal/mol) relative to the native ligand&#xa0;(-9.56 kcal/mol) against the FGFR1 receptor protein. At the same time, MDS revealed&#xa0;that it remained firmly bound at the catalytic site of the protein throughout the 200 ns&#xa0;simulation run, with minimal variations in root-mean-square fluctuation (RMSF) and&#xa0;root-mean-square deviation (RMSD). Binding free energy changes (-29.84 ±&#xa0;3.35 kcal/mol) further validated the thermodynamic stability of protein-M6 adduct and&#xa0;spontaneity of its formation.&#xa0;</p> Conclusion <p>These findings underscore the effectiveness of ML-driven virtual screening,&#xa0;molecular docking, and MDS in accelerating the discovery of novel FGFR1 inhibitors with potential anticancer properties. The hit candidate is recommended for further experimental assays.</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

Machine Learning-guided Virtual Screening and Molecular Dynamics Simulations for Identifying Potent FGFR1 Inhibitors as Anti-neoplastic Agents

  • 期刊论文

  • 发表日期：

    2026年3月6日

查看全文
  • Ram Lal Swagat Shrestha,
  • Manila Poudel,
  • Sandeep Poudel Chhetri,
  • Ashika Tamang,
  • Nirmal Parajuli,
  • Shiva M.C.,
  • Aakar Shrestha,
  • Timila Shrestha,
  • Samjhana Bharati,
  • Binita Maharjan,
  • Bishnu P. Marasini,
  • Jhashanath Adhikari Subin

摘要

Purpose

Quick identification of new molecules against protein involved in cancer pathogenesis by machine learning (ML) approach is the main goal of this research. The fibroblast growth factor receptor 1 (FGFR1) is a robust target in cancer therapy, with its overexpression linked to various malignancies. Despite the availability of FDA approved FGFR1 inhibitors, drug resistance and specificity issues highlight the need for novel therapeutic candidates.

Methods

This study integrates ML with molecular docking and molecular dynamics simulations (MDS) to identify and validate potential FGFR1 inhibitors.

Results

ML models were trained and optimized to build a voting classifier using 1,523 data points of FGFR1 known inhibitors, attaining an accuracy of 89% and an AUC score (area under the curve) of 0.95. The optimized model was applied to COCONUT database (695,133 molecules), yielding 12 potential inhibitors, among which ligand M6 exhibited the highest predicted activity. Molecular docking confirmed its stronger binding affinity (-9.55 kcal/mol) relative to the native ligand (-9.56 kcal/mol) against the FGFR1 receptor protein. At the same time, MDS revealed that it remained firmly bound at the catalytic site of the protein throughout the 200 ns simulation run, with minimal variations in root-mean-square fluctuation (RMSF) and root-mean-square deviation (RMSD). Binding free energy changes (-29.84 ± 3.35 kcal/mol) further validated the thermodynamic stability of protein-M6 adduct and spontaneity of its formation. 

Conclusion

These findings underscore the effectiveness of ML-driven virtual screening, molecular docking, and MDS in accelerating the discovery of novel FGFR1 inhibitors with potential anticancer properties. The hit candidate is recommended for further experimental assays.