Comparison of various methods for validity evaluation of QSAR models

Abstract

Background Quantitative structure–activity relationship (QSAR) modeling is one of the most important computational tools employed in drug discovery and development. The external validation of QSAR models is the main point to check the reliability of developed models for the prediction activity of not yet synthesized compounds. It was performed by different criteria in the literature. Methods In this study, 44 reported QSAR models for biologically active compounds reported in scientific papers were collected. Various statistical parameters of external validation of a QSAR model were calculated, and the results were discussed. Results The findings revealed that employing the coefficient of determination (r2) alone could not indicate the validity of a QSAR model. The established criteria for external validation have some advantages and disadvantages which should be considered in QSAR studies. Conclusion This study showed that these methods alone are not only enough to indicate the validity/invalidity of a QSAR model.

The development and impact of an app for a smart drug interaction reminder system

“…This study’s main algorithm was a basic artificial neural network. The results can be combined with the design concept of QSAR in future research [ 30 ]. More machine learning algorithms can be incorporated, such as binary logistic regression and random forest, to calculate the possible side effects of medication for specific patients.…”

Innovative strategies for the quantitative modeling of blood–brain barrier (BBB) permeability: harnessing the power of machine learning-based q-RASAR approach

“…Acknowledging the limitations of internal validation metrics in assessing the predictive accuracy, our study undertook a comprehensive validation of test set compounds, employing various external validation metrics encompassing Q F1 2 , Q F2 2 , MAE test , r m 2 metrics such as average r m 2 (test), Δ r m 2 (test), and concordance correlation coefficient (CCC). 57–63 The utilization of these external validation methods fortifies the claim of the predictive nature of the developed model. Furthermore, we conducted the Y -randomization test, 64 applicability domain criteria (DModX-distance to model in the X -space), etc.…”

“…Internal statistical quality and validation metrics, including determination coefficient (R 2 ), leave-one-out cross-validated correlation coefficient (Q (LOO) concordance correlation coefficient (CCC). [57][58][59][60][61][62][63] The utilization of these external validation methods fortifies the claim of the predictive nature of the developed model. Furthermore, we conducted the Y-randomization test, 64 applicability domain criteria (DModX-distance to model in the X-space), etc., utilizing Simca-P 10.0 software.…”

Recent Advances in Machine-Learning-Based Chemoinformatics: A Comprehensive Review

“…External validation is the primary method for evaluating the accuracy of generated models for the activity prediction of compounds that have not yet been synthesized. Understanding the variables that control molecular characteristics and creating new compounds with advantageous features depend on QSAR models, which provide information on the association between activities and structure-based molecular descriptors [96,97].…”