Logo image
Back
MoleculeNet: a benchmark for molecular machine learning
Journal article   Open access

MoleculeNet: a benchmark for molecular machine learning

Zhenqin Wu, Bharath Ramsundar, Evan N Feinberg, Joseph Gomes, Caleb Geniesse, Aneesh S Pappu, Karl Leswing and Vijay Pande
Chemical science (Cambridge), Vol.9(2), pp.513-530
01/14/2018
DOI: 10.1039/c7sc02664a
PMID: 29629118
url
https://doi.org/10.1039/c7sc02664aView
Published (Version of record) Open Access

Abstract

Molecular machine learning has been maturing rapidly over the last few years. Improved methods and the presence of larger datasets have enabled machine learning algorithms to make increasingly accurate predictions about molecular properties. However, algorithmic progress has been limited due to the lack of a standard benchmark to compare the efficacy of proposed methods; most new algorithms are benchmarked on different datasets making it challenging to gauge the quality of proposed methods. This work introduces MoleculeNet, a large scale benchmark for molecular machine learning. MoleculeNet curates multiple public datasets, establishes metrics for evaluation, and offers high quality open-source implementations of multiple previously proposed molecular featurization and learning algorithms (released as part of the DeepChem open source library). MoleculeNet benchmarks demonstrate that learnable representations are powerful tools for molecular machine learning and broadly offer the best performance. However, this result comes with caveats. Learnable representations still struggle to deal with complex tasks under data scarcity and highly imbalanced classification. For quantum mechanical and biophysical datasets, the use of physics-aware featurizations can be more important than choice of particular learning algorithm.

Details

Metrics

142 Record Views
2075 Times Cited - Web of Science
Logo image