Abstract
The enzyme kinase is a member of a broad family that catalyses the transfer of highly energetic phosphate molecules to substrates like protein, lipids, carbohydrates, and nucleic acid. Protein tyrosine kinase is a target for therapeutic intervention since it is crucial for several immunological and signal transduction processes. Protein kinase is dysregulated, overexpressed, and mutated in a variety of disorders, including cancer and immunopathological conditions. Drug discovery techniques developed in-silico are far more affordable and quicker than those used now. The utility of the QSAR model in the current study is demonstrated in the search for novel tyrosine kinase inhibitors. 100 highly powerful compounds were chosen out of a total of 7226 compounds that were pulled from the ChEMBL database. For each chemical, more than 2000 descriptors of various classes were computed. Techniques for feature selection and outlier reduction were employed to cut down on the quantity of unimportant characteristics. The final QSAR model is created using the SVR, MLR, RF, and RT machine learning techniques. We also used the internal and external assessment criteria to assess the model's predictability and stability. For the training set, the four created models all displayed acceptable R2 values of 85, 81, 91, and 94 for MLR, SVR, RF, and RT, respectively. Apart from the RT technique, the test dataset's evaluation used the same matrix and showed virtually identical values to those of the train set. A Y-randomization test was also conducted, and the results showed that the model was not generated randomly.
Keyword
Protein tyrosine kinase, QSAR, Cancer, Machine learning, MLR, SVR, RF, RT
PDF Download (click here)
|