TABLE 4.
Examples of key AI-driven algorithms and methods for prediction of molecular properties.
| Class | Method description | References/Examples |
|---|---|---|
| Dynamic Selection | Techniques for dynamic selection of classifiers based on individual sample | Cruz et al. (2018) |
| Ensemble learning | Combination of multiple learners for performance improvement | Svetnik et al. (2003); Sheridan et al. (2016); Kwon et al. (2019); Davronov and Adilova, (2021) |
| Fully Connected DL | Fully connected deep learning network for Single-task QSAR analysis | Ma et al. (2015) |
| GCN | Multitask graph convolutional networks | Montanari et al. (2019) |
| Fully Connected DL | Fully connected deep learning network for Multi-task QSAR analysis | Kearnes et al. (2016) |
| GCN | PotentialNet family of graph convolutions for protein-ligand binding affinity | Feinberg et al. (2018); Feinberg et al. (2020) |
| GNN | Molecular Contrastive Learning | Wang et al. (2021c) |
| MPNN | Message passing neural networks for molecular property prediction | (Yang X. et al. (2019); Stokes et al. (2020); Chen et al. (2021); Heid et al. (2023) |
| Graph Transformer | Molecular encoding using hybrid MPNN-Transformer architectures | Rong et al. (2021) |
| NLP inspired | Autoencoder-based Molecular encoding and QSAR | Winter et al. (2019) |
| NLP inspired | Transformer based encoder model | Honda et al. (2019); Payne et al. (2020); Irwin, Dimitriadis et al. (2022) |
| NLP inspired | Transfer learning for NLP based classification tasks | Li and Fourches (2020) |
| NLP inspired | Siamese RNNs for QSAR Prediction | Fernández-Llaneza et al. (2021) |
| Active Learning | Retrosynthetic and combinatorial synthesis coupled with Active Learning | Konze et al. (2019) |
| Explainable Artificial Intelligence | Methods to provide interpretability to ML/DL models. These include approaches for explaining their predictions, quantifying their uncertainty, and estimating their applicability domains | Interpretability Ribeiro et al. (2016); Lundberg and Lee, (2017); Nori et al. (2019); Rodríguez-Pérez and Bajorath, (2021) |
| Uncertainty estimation Liu et al. (2018a); Cortés-Ciriano and Bender, (2019); Gawlikowski et al. (2021); Zhong et al. (2022) | ||
| Applicability domain Liu and Wallqvist, (2019), R. P. Sheridan, (2015); Schroeter et al. (2007); Supratik et al. (2018) |