Is regression through origin useful in external validation of QSAR models?

Area of Science:

• * Computational Chemistry
• * Toxicology
• * Drug Discovery

Background:

• * Quantitative Structure-Activity Relationship (QSAR) models are vital for predicting chemical properties and assessing new chemicals.
• * Current external validation methods, like Golbraikh-Tropsha and Roy, rely on Regression Through Origin (RTO) criteria.
• * Inconsistencies in RTO's r(2) calculation raise concerns about their reliability.

Purpose of the Study:

• * To evaluate the reliability of existing QSAR external validation methods.
• * To identify limitations in Regression Through Origin (RTO) based validation criteria.
• * To propose a more robust approach for QSAR model external validation.

Main Methods:

• * Calculation of deviation parameters, including absolute errors, to assess model performance.
• * Comparison of errors between training and test datasets.
• * Analysis of inconsistencies in the definition and calculation of r(2) for RTO.

Main Results:

• * Existing RTO-based validation criteria showed poor agreement with deviation parameter calculations.
• * Inconsistencies were found in the definition and calculation of r(2) for RTO.
• * The current RTO-based criteria are not optimal for external validation.

Conclusions:

• * The standard RTO-based criteria for QSAR external validation are suboptimal.
• * Comparing model errors on training and test sets presents a more reliable validation approach.
• * This error-comparison method enhances the trustworthiness of QSAR models for new chemical assessments.