Predicting EGFR^{L858R/T790M/C797S} Inhibitory Effect of Osimertinib Derivatives by Mixed Kernel SVM Enhanced with CLPSO

Area of Science:

• Computational chemistry
• Drug discovery
• Oncology

Background:

• Epidermal growth factor receptor (EGFR) mutations, specifically EGFR^{L858R/T790M/C797S}, confer resistance to Osimertinib.
• Developing novel inhibitors is crucial to overcome Osimertinib resistance in cancer treatment.

Purpose of the Study:

• To predict the inhibitory effects of Osimertinib derivatives against EGFR^{L858R/T790M/C797S} mutations.
• To guide the design and screening of more effective EGFR inhibitors.

Main Methods:

• Six predictive models were developed, including heuristic method (HM), random forest (RF), gene expression programming (GEP), gradient boosting decision tree (GBDT), and two support vector machine (SVM) variants (polynomial and mixed kernel).
• Model descriptors were selected using heuristic methods or XGBoost, with hyperparameters optimized by a comprehensive learning particle swarm optimizer.
• Internal and external validation employed leave-one-out cross-validation (QLOO2), 5-fold cross-validation (Q5-fold2), concordance correlation coefficient (CCC), QF12, and QF22.
• Molecular docking analysis explored properties of novel EGFR inhibitors.

Main Results:

• The mixed kernel SVM (MIX-SVM) model demonstrated superior performance, achieving high R2 (0.9445 training, 0.9490 test) and low RMSE (0.1659 training, 0.1814 test).
• Excellent validation metrics were obtained: QLOO2 (0.9107), Q5-fold2 (0.8621), CCC (0.9835), QF12 (0.9689), and QF22 (0.9680).
• The HM model predicted IC50 values for 162 novel compounds, with top candidates validated via PEA.

Conclusions:

• The MIX-SVM model provides a robust platform for predicting Osimertinib derivative efficacy.
• This approach offers valuable guidance for the rational design and efficient screening of novel EGFR^{L858R/T790M/C797S} inhibitors.
• The findings contribute to the development of next-generation EGFR inhibitors for resistant cancers.