Gaussian Process Molecule Property Prediction with FlowMO
This work addresses the problem of predicting molecular properties with reliable uncertainty estimates for researchers in chemistry and drug discovery, offering an incremental improvement through better calibration in small-data scenarios.
The authors tackled molecular property prediction by developing FlowMO, an open-source Python library using Gaussian Processes, which achieved comparable predictive performance to deep learning methods on three small datasets while providing superior uncertainty calibration.
We present FlowMO: an open-source Python library for molecular property prediction with Gaussian Processes. Built upon GPflow and RDKit, FlowMO enables the user to make predictions with well-calibrated uncertainty estimates, an output central to active learning and molecular design applications. Gaussian Processes are particularly attractive for modelling small molecular datasets, a characteristic of many real-world virtual screening campaigns where high-quality experimental data is scarce. Computational experiments across three small datasets demonstrate comparable predictive performance to deep learning methods but with superior uncertainty calibration.