Neural Message Passing for Quantum Chemistry

Supervised learning on molecules has incredible potential to be useful in chemistry, drug discovery, and materials science. Luckily, several promising and closely related neural network models invariant to molecular symmetries have already been described in the literature... These models learn a message passing algorithm and aggregation procedure to compute a function of their entire input graph. At this point, the next step is to find a particularly effective variant of this general approach and apply it to chemical prediction benchmarks until we either solve them or reach the limits of the approach. In this paper, we reformulate existing models into a single common framework we call Message Passing Neural Networks (MPNNs) and explore additional novel variations within this framework. Using MPNNs we demonstrate state of the art results on an important molecular property prediction benchmark; these results are strong enough that we believe future work should focus on datasets with larger molecules or more accurate ground truth labels. read more

PDF Abstract ICML 2017 PDF ICML 2017 Abstract

Results from the Paper


Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Node Classification CiteSeer (0.5%) MPNN Accuracy 41.8% # 13
Node Classification CiteSeer (1%) MPNN Accuracy 54.3% # 12
Node Classification CiteSeer with Public Split: fixed 20 nodes per class MPNN Accuracy 64.0% # 35
Node Classification Cora (0.5%) MPNN Accuracy 46.5% # 13
Node Classification Cora (1%) MPNN Accuracy 56.7% # 13
Node Classification Cora (3%) MPNN Accuracy 72.0% # 12
Node Classification Cora with Public Split: fixed 20 nodes per class MPNN Accuracy 78.0% # 31
Graph Regression Lipophilicity MPNN RMSE 0.719 # 5
Node Classification PubMed (0.03%) MPNN Accuracy 53.9% # 12
Node Classification PubMed (0.05%) MPNN Accuracy 59.6% # 12
Node Classification PubMed (0.1%) MPNN Accuracy 67.3% # 12
Node Classification PubMed with Public Split: fixed 20 nodes per class MPNN Accuracy 75.6% # 32
Drug Discovery QM9 MPNNs Error ratio 0.68 # 4
Formation Energy QM9 MPNN MAE 0.45 # 13
Graph Regression ZINC 100k MPNN MAE 0.288 # 3
Graph Regression ZINC-500k MPNN (max) MAE 0.252 # 13
Graph Regression ZINC-500k MPNN (sum) MAE 0.145 # 8

Methods