BI433: SM-GATC : Predicting The Efficacy Of Thai Herbs For Inhibitory EGFR

Nichakorn Intharapradit Varee Chiang Mai School

MIIX24 | Beginner Innovator

CR: 0.0000 | 0 Likes | 4 Views | 61 times | LS: 61.0
Like it? | Support them now!

Currently, there are numerous options for cancer treatment, but the process of discovering and producing drugs often involves extensive time and high costs. Through research, it has been found that various types of Thai herbal substances possess medicinal properties and potential as drugs. In the medical field, researchers have initiated this project with the aim of developing an artificial intelligence model to predict the inhibitory capabilities against EGFR protein of compounds found in Thai herbs.The project involves several steps, starting with the collection of biological data on substances targeting EGFR from the ChEMBL database, totaling 18,762 entries. After data refinement, a dataset of 2,151 entries was obtained. A machine learning model was then constructed using techniques such as transforming the SMILES structure into fingerprints, molecular descriptors, and graph representation of atom and bond properties. Graph Attention Network and Convolutional Neural Network were employed to identify crucial features of the compounds.The performance of the our model (SM-GATC) was evaluated through experiments, resulting in a Root Mean Squared Error (RMSE) of 0.46 and an R-square value of 0.93. Testing on a separate dataset (Test dataset) consisting of molecular structures of 130 cancer treatment drugs, resulting in an RMSE of 0 . 48 and an R-square value of 0 . 92 When predicting pIC5 0 values for herbal compounds in coriander and cannabis.The results of this experiment can be valuable for drug discovery and development, specifically in the context of cancer treatment in Thailand.