Calibration and generalizability of probabilistic models on low-data chemical datasets with DIONYSUS
This work addresses the challenge of model selection and calibration for researchers in chemistry dealing with limited data, though it is incremental as it focuses on benchmarking existing methods.
The authors tackled the problem of evaluating probabilistic machine learning models on small chemical datasets (<2000 molecules), finding that deep learning may not be optimal and providing practical insights into model and feature choices for tasks like binary classification and regression.
Deep learning models that leverage large datasets are often the state of the art for modelling molecular properties. When the datasets are smaller (< 2000 molecules), it is not clear that deep learning approaches are the right modelling tool. In this work we perform an extensive study of the calibration and generalizability of probabilistic machine learning models on small chemical datasets. Using different molecular representations and models, we analyse the quality of their predictions and uncertainties in a variety of tasks (binary, regression) and datasets. We also introduce two simulated experiments that evaluate their performance: (1) Bayesian optimization guided molecular design, (2) inference on out-of-distribution data via ablated cluster splits. We offer practical insights into model and feature choice for modelling small chemical datasets, a common scenario in new chemical experiments. We have packaged our analysis into the DIONYSUS repository, which is open sourced to aid in reproducibility and extension to new datasets.