Implications of Additivity and Nonadditivity for Machine Learning and Deep Learning Models in Drug Design

Abstract

Matched molecular pairs (MMPs) is nowadays a commonly applied concept in drug design. It is used in many computational tools for structure activity relationship analysis, biological activity prediction or optimization of physicochemical properties. However, up to date it has not been shown in a rigorous way that MMPs, i.e. changing only one substituent between two molecules, can be predicted with high accuracy and precision in contrast to any other chemical compound pair. It is expected that any model should be able to predict such a de ned change with high accuracy and reasonable precision. In this study, we examine the predictability of four classical properties relevant for drug design ranging from simple physicochemical parameters (logD and solubility) to more complex cell based ones (permeability and clearance), using different data sets and machine learning algorithms.Our study con rms that additive data is the easiest to predict which highlights the importance of recognition of nonadditivity events and the challenging complexity of predicting properties in case of scaffold hopping. Despite of deep learning being well suited to model non-linear events, these methods do not seem to be an exception of this observation. Though, they are in general performing better than classical machine learning methods, this leaves the eld with a still standing challenge.

Exposing the Limitations of Molecular Machine Learning with Activity Cliffs

“…Similar results were obtained when comparing graph networks for (a) feature attribution with activity cliffs, 89 and (b) bioactivity prediction. 30 A recent analysis on physicochemical-property cliffs highlights an opposite trend, with deep learning methods performing better than simpler machine learning approaches 90 —potentially due to the higher number of training samples (approx. 20,000 molecules).…”

Advancements in Ligand-Based Virtual Screening through the Synergistic Integration of Graph Neural Networks and Expert-Crafted Descriptors

“…Recently, some authors expressed doubts about deep learning performance over traditional methods in molecular tasks [14,15]. Ultimately, it remains a mystery whether GNNs are consistently better than methods that rely on traditional descriptor in CADD [16,17]. We hypothesize that these conflicting reports might invite an integrated method that combined GNNs with traditional descriptors to outperform both individual approaches, at least at the moment.…”

Generative artificial intelligence in drug discovery: basic framework, recent advances, challenges, and opportunities

“…The importance of conditional generation algorithms is predicted to grow in the coming years. These techniques may enable the generation of molecules designed to meet specific requirements, potentially overcoming the limits of current scoring systems (for instance, because of non-additivity, activity cliffs ( Kwapien et al, 2022 ; Özçelik et al, 2022 ). A promising structure-based design may address de novo design for as-yet-undiscovered macromolecular targets ( Volkov et al, 2022 ), by producing molecules that match specific binding sites’ electrostatic and shape properties.…”