I have obtained a set of 200 selected Mordred descriptors calculated from 3D structures optimized at the semi empirical PM7 level of theory with implicit water solvation using the COSMO water colvation model from @IamDavyG and @luiraym. I used Weka 3.8.3 to select attributes based on an unsupervised standardisation approach and then correlate properties with potency outcomes. Based on performance in 10-fold CV I selected a Random Forest model which produced the lowest RMSE in training. The final RMSE was 0.805 mean absolute error was 0.666.
I have obtained a set of 200 selected Mordred descriptors calculated from 3D structures optimized at the semi empirical PM7 level of theory with implicit water solvation using the COSMO water colvation model from @IamDavyG and @luiraym. I used Weka 3.8.3 to select attributes based on an unsupervised standardisation approach and then correlate properties with potency outcomes. Based on performance in 10-fold CV I selected a Random Forest model which produced the lowest RMSE in training. The final RMSE was 0.805 mean absolute error was 0.666.