molecular-conformer-generation

Benchmark datasets

The benchmark datasets in molecular conformation generation task are GEOM-QM9 and GEOM-Drugs.

The detail information of the benchmark is referred from "Zhu J , Xia Y , Liu C , et al. Direct Molecular Conformation Generation[J]. arXiv e-prints, 2022":

	GEOM-QM9	GEOM-Drugs
Training set	200K	200K
Validation set	2.5K	2.5K
Test set	22408	14324

large-scale setting

	GEOM-QM9	GEOM-Drugs
Training set	1.37M	2M
Validation set	165K	100K
Test set	174K	100K

Datasets

Reviews

Progress in Computational Chemistry (Interdisciplinary Science Letters, 2018) [Paper]

Direct generation methods

[CoarseGrainingVAE] Generative Coarse-Graining of Molecular Conformations (arXiv, 2022) [Paper] [Code]

[DMCG] Direct Molecular Conformation Generation (arXiv, 2022) [Paper] [Code]

[GEODIFF] GeoDiff: a Geometric Diffusion Model for Molecular Conformation Generation (ICLR, 2022) [Paper] [Code]

[CGVAE] Generative Coarse-Graining of Molecular Conformations (ICML, 2022) [Paper] [Code]

Intermediate geometric elements-based methods

Intermediate geometric elements include atomic distances, torsion angles and the gradients w.r.t. (with regard to) inter-atomic distances, etc.

Inter-atomic distances

[CGCF-ConfGen] Learning Neural Generative Dynamics for Molecular Conformation Generation (ICLR, 2021) [Paper] [Code]

[GraphDG] A generative model for molecular distance geometry (ICML, 2020) [Paper] [Code]

[GraphAF] GraphAF: a Flow-based Autoregressive Model for Molecular Graph Generation (ICLR, 2020) [Paper] [Code]

Torsion angles

The gradients w.r.t inter-atomic distances

[ConfGF] Learning Gradient Fields for Molecular Conformation Generation (ICML, 2021) [Paper] [Code]

[DGSM] Predicting Molecular Conformation via Dynamic Graph Score Matching (NeurIPS, 2021) [Paper] [Code]