MoleculeNet

Zhenqin Wu, Bharath Ramsundar, Evan N. Feinberg, Joseph Gomes, Caleb Geniesse, Aneesh S. Pappu, Karl Leswing, Vijay Pande

Abstract

A large scale benchmark for molecular machine learning consisting of multiple public datasets, metrics, featurizations and learning algorithms.

Tested on 3 datasets

Tables show presented results in global context (best 3 papers + GCN + MLP).

#	Model	RMSE ↓	Paper
1	Graphormer-SPIS code ↗ GT global	0.4840 ±0.0050	On Structural Expressive Power of Graph Transformers code ↗
2	Graphormer-SPD code ↗ GT global	0.4920 ±0.0040	On Structural Expressive Power of Graph Transformers code ↗
3	conan-FGW global	0.5140 ±0.0190	Structure-Aware E(3)-Invariant Molecular Conformer Aggregation Networks
7	GCN code ↗ GNN global	0.5640 ±0.0150	On Structural Expressive Power of Graph Transformers code ↗
12	MPNN GNN baseline	0.5800 ±0.0300	This paper
20	Weave GNN baseline	0.6100 ±0.0700	This paper
50	DAG GNN baseline	0.8200 ±0.0800	This paper
54	XGBoost Boost baseline actual #54 vs paper #84 30 places worse	0.8320 best reported	actual #54 vs paper #84 30 places worse paper 0.9900 ±0.1400 best today: BiScale-GTR: Fragment-Aware Graph Transformers for Multi-Scale Molecular Representation Learning (2026) Difference within reported uncertainty; not counted as underperformance. paper 0.9900 ±0.1400 Δ today 0.1580
70	rf RF baseline actual #70 vs paper #93 23 places worse	0.9180 best reported	actual #70 vs paper #93 23 places worse paper 1.0700 ±0.1900 best today: BiScale-GTR: Fragment-Aware Graph Transformers for Multi-Scale Molecular Representation Learning (2026) Difference within reported uncertainty; not counted as underperformance. paper 1.0700 ±0.1900 Δ today 0.1520
80	GC GNN baseline	0.9700 ±0.0100	This paper
106	Multitask GNN baseline	1.1200 ±0.1500	This paper
133	KRR Trad baseline	1.5300 ±0.0600	This paper

#	Model	RMSE ↓	Paper
1	SE(3)-I MPPH Hyb global	0.3540 ±0.0530	SE(3)-Invariant Multiparameter Persistent Homology for Chiral-Sensitive Molecular Property …
2	ECFP-4 + RF RF global	0.5600 ±0.0660	SE(3)-Invariant Multiparameter Persistent Homology for Chiral-Sensitive Molecular Property …
3	GPSE + GPS code ↗ ext data Hyb global	0.6820	Graph Positional and Structural Encoder code ↗
17	MPNN GNN baseline	1.1500 ±0.1200	This paper
19	Weave GNN baseline	1.2200 ±0.2800	This paper
22	GC GNN baseline	1.4000 ±0.1600	This paper
32	DAG GNN baseline	1.6300 ±0.1800	This paper
38	XGBoost Boost baseline	1.7400 ±0.1500	This paper
42	Multitask GNN baseline	1.8700 ±0.0700	This paper
50	rf RF baseline	2.0300 ±0.2200	This paper
57	KRR Trad baseline	2.1100 ±0.0700	This paper
69	GIN code ↗ GNN global	2.3460 ±0.1220	ChemRL-GEM: Geometry Enhanced Molecular Representation Learning for Property Prediction code ↗

#	Model	RMSE ↓	Paper
1	MemGNN code ↗ Hyb global	0.5560 ±0.0230	Memory-Based Graph Networks code ↗
2	ProtoS-L2 Hyb 0.1M global	0.5800 ±0.0160	Optimal Transport Graph Neural Networks
3	ProtoW-L2 Hyb global	0.6040 ±0.0140	Optimal Transport Graph Neural Networks
8	GC GNN baseline	0.6550 ±0.0360	This paper
7	GCN code ↗ GNN global	0.6550 ±0.0360	Memory-Based Graph Networks code ↗
22	MPNN GNN baseline actual #22 vs paper #27 5 places worse	0.7120 best reported	actual #22 vs paper #27 5 places worse paper 0.7190 ±0.0310 best today: Graph Positional and Structural Encoder (ICML 2023) Difference within reported uncertainty; not counted as underperformance. paper 0.7190 ±0.0310 Δ today 0.0070
26	Weave GNN baseline	0.7150 ±0.0350	This paper
57	XGBoost Boost baseline	0.7990 ±0.0540	This paper
65	DAG GNN baseline	0.8350 ±0.0390	This paper
72	Multitask GNN baseline	0.8590 ±0.0130	This paper
75	rf RF baseline	0.8760 ±0.0400	This paper
80	KRR Trad baseline	0.8990 ±0.0430	This paper

SOTA performance (not significantly different from the best) Baseline / comparison (this paper) Global context (other papers)