Drug Discovery Analysis Using Machine Learning Bioinformatics

Abstract

Bioinformatics is defined as the application of tools, computation, analysis to capture the bioactivity data and interpretation of biological data. To develop or to discover the drug, the biological information of proteins, cells, RNA’S, DNA is required to analyze those functional behavior of compounds structures, physical and chemical properties. For example: In the process of discovering a new protein sequence, it utilizes the known/existing sequences in order to compare the similarities or features of the newly discovered sequence. By looking, we can roughly tell about the functioning of the newly discovered sequence of proteins. Drugs are one of the chemical substances that can change the way our body and mind works. Discovery of drugs aims to find the compounds that are classified based on the threshold values to be active or an inactive and then analyzes drug-likeness properties of compounds like absorption, distribution, metabolism and excretion (ADME). Then, we compared the molecular features and properties of each and every compound once the salts, impurities, organic acids are removed from the molecular structures. This comparison helps in preparing the dataset with PubChem fingerprints and pIC50 values as X (input variables) and Y (output variables). In the proposed system, decision tree regressor got good performance in its R-squared value and got the very lowest RMSE. Hence, the time taken for the evaluation is also very very low. These values represent that the proposed dataset is fit for regression prediction based on values achieved in work. Finally, the scatter plot is builded in between the experimental and the predicted values of pIC50 to get the line of regression. In general, this application can be helpful for doctors and researchers in developing a drug.