English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Deep learning models that leverage large datasets are often the state of theart for modelling molecular properties. When the datasets are smaller (< 2000molecules), it is not clear that deep learning approaches are the rightmodelling tool. In this work we perform an extensive study of the calibrationand generalizability of probabilistic machine learning models on small chemicaldatasets. Using different molecular representations and models, we analyse thequality of their predictions and uncertainties in a variety of tasks (binary,regression) and datasets. We also introduce two simulated experiments thatevaluate their performance: (1) Bayesian optimization guided molecular design,(2) inference on out-of-distribution data via ablated cluster splits. We offerpractical insights into model and feature choice for modelling small chemicaldatasets, a common scenario in new chemical experiments. We have packaged ouranalysis into the DIONYSUS repository, which is open sourced to aid inreproducibility and extension to new datasets.

Calibration and generalizability of probabilistic models on low-data chemical datasets with DIONYSUS