Quantitative structure-activity relationship study of skin sensitization of Michael acceptors based on quantum chemical descriptors

Read-across-based intelligent learning: development of a global q-RASAR model for the efficient quantitative predictions of skin sensitization potential of diverse organic chemicals

“…The developed PLS q-RASAR model superseded the corresponding PLS QSAR model both in terms of internal and external validation metric values. Most of the previous works 68,69 have performed QSAR analysis based on chemicals that share the same mechanism of skin sensitization, and thus, lack diversity in their datasets. Moreover, the number of compounds required to train and validate their models was also quite low.…”

“…In spite of considering diverse organic chemicals, which have different mechanisms of skin sensitization, our global PLS q-RASAR model is robust and also possesses good external predictivity as evident from the different internal and external validation metrics. Manhas et al 69 performed QSAR analysis using a small dataset ( n = 30) consisting of compounds that are Michael acceptors. From their internal validation statistics, we can infer that our PLS q-RASAR model is much more robust since our cross-validated Q 2 is much higher with a similar value of R 2 despite the fact that our model has been generated on a much larger number of compounds with structurally and mechanistically diverse organic compounds.…”