Applicability domain

Applicability domain refers to the range of conditions and limitations within which a model can be reliably applied. It is a critical factor in determining the accuracy and validity of a model, as it defines the scope of its predictions and identifies potential sources of uncertainty. The process of defining the applicability domain involves assessing the model's performance under different conditions and establishing the boundaries of its predictive capabilities.From: A review on machine learning methods for in silico toxicity prediction [2018], Extrapolation Practice for Ecotoxicological Effect Characterization of Chemicals [2008], Nanocatalysis [2019]

Nano-QSAR for Environmental Hazard Assessment: Turning Challenges into Opportunities

View Chapter

Purchase Book

Published in Agnieszka Gajewicz, Tomasz Puzyn, Computational Nanotoxicology, 2019

Willie Peijnenburg, Guangchao Chen, Vijver Martina

where P(N P ∈ T) and P(N P ∈ N) are the probabilities of an ENM being classified as toxic or nontoxic, respectively, d is the ENM size, θv is the volume concentration derived from the mass concentration of ENMs, PMe is the period of the ENM metal in the periodic table, and EMeO is the atomization energy of the metal oxide. The model applicability domain was depicted by the principal component analysis.

A review on machine learning methods for in silico toxicity prediction

View Article

Journal Information

Published in Journal of Environmental Science and Health, Part C, 2018

Gabriel Idakwo, Joseph Luttrell, Minjun Chen, Huixiao Hong, Zhaoxian Zhou, Ping Gong, Chaoyang Zhang

Regardless of how generalized a model may appear to be following validation, it is impractical to consider the model applicable to the entire chemical space. The predictions made by models on new compounds with descriptor values outside the training data descriptor (feature) space may not be reliable. It is therefore necessary to know the boundary within which the model can extrapolate reliably. The applicability domain (AD) defines the scope and limitations of a model. AD attempts to define the degree of generalization of the model by highlighting the range of chemical structures for which the model is considered to be reliably applicable.94,95 Predictions of compounds outside a model’s AD cannot be considered reliable.

Applicability domain

Explore chapters and articles related to this topic

Nano-QSAR for Environmental Hazard Assessment: Turning Challenges into Opportunities

A review on machine learning methods for in silico toxicity prediction