Integrating machine learning and structure-based approaches for repurposing potent tyrosine protein kinase Src...

Area of Science:

• Biochemistry and Molecular Biology
• Computational Chemistry and Cheminformatics
• Pharmacology and Drug Discovery

Background:

• Tyrosine-protein kinase Src is crucial for cell growth but implicated in inflammatory diseases upon overexpression or mutation.
• Existing Src inhibitors like dasatinib face specificity and selectivity challenges.
• Novel, targeted, and non-toxic inhibitors are needed for effective treatment of Src-related inflammatory conditions.

Purpose of the Study:

• To identify novel, targeted, and non-toxic Src kinase inhibitors using an integrated machine learning and structure-based drug repurposing strategy.
• To screen FDA-approved drugs for potential repurposing as Src kinase inhibitors.
• To evaluate the efficacy and safety of identified drug candidates for treating inflammatory diseases.

Main Methods:

• Machine learning models (SVM, RF, K-NN, Decision Tree) were trained on existing bioactivity data to predict Src kinase inhibition.
• A library of 1040 FDA-approved drugs was screened using the best-performing SVM model.
• Molecular docking, molecular dynamics simulations, and MMGBSA analysis were used to assess binding affinity, stability, and interactions.
• In silico toxicity analysis was performed to evaluate potential safety concerns.

Main Results:

• Support Vector Machine (SVM) was identified as the optimal machine learning model for predicting compound bioactivity.
• 51 potent Src kinase inhibitor candidates were shortlisted from the FDA-approved library.
• Orlistat, acarbose, and afatinib emerged as leading candidates with stable conformations and strong binding interactions.
• MMGBSA analysis indicated favorable binding free energies for orlistat, acarbose, and afatinib compared to dasatinib.
• Orlistat and acarbose were identified as potentially safer therapeutics due to lower predicted toxicity compared to afatinib.

Conclusions:

• Integrated computational approaches, including machine learning and structure-based drug repurposing, are effective for identifying novel drug candidates.
• Orlistat and acarbose represent promising candidates for further experimental validation as Src kinase inhibitors for inflammatory diseases.
• The study highlights the potential of computational methods to accelerate drug discovery and develop safer therapeutics for Src-related conditions.