Graph Diffusion Models: A Comprehensive Survey of Methods and Applications

This is the summation of all the methods, datasets, and other survey mentioned in our survey 'Graph Diffusion Models: A Comprehensive Survey of Methods and Applications' 🔥. Any problems, please contact [email protected]. Any other interesting papers or codes are welcome. If you find this repository useful to your research or work, it is really appreciated to star this repository ❤️.

Contents

• Diffusion-based Graph Generative Methods
  • Contents
  • Molecule generation
    • Molecule design
    • Conformation design
    • De novo ligand design
    • Ligand docking
    • Protein design
  • Motion generation
    • Motion synthesis
    • Motion prediction
  • Others
  • Datasets
    • Molecule generation
    • Motion generation
  • Other surveys

Molecule generation

De novo molecule design

Methods	Paper	Code	Methods	Paper	Code
DiGress (ICLR-23)	[paper]	[code]	MiDi (ICLR-23)	[paper]	[code]
CDGS (NeurIPS-22)	[paper]	[code]	GCDM (ICLR-23)	[paper]	[code]
EDM (ICML-22)	[paper]	[code]	Wu et al. (NeurIPS-22)	[paper]	-
MDM (AAAI-23)	[paper]	[code]	DiffLinker	[paper]	[code]
JODO	[paper]	[code]	SILVR	[paper]	-
HierDiff (ICML-23)	[paper]	[code]	-	-	-
BIMODAL	[paper]	[code]	RationaleRL (ICML20)	[paper]	[code]
GEOLDM (ICML-23)	[paper]	[code]	MGM	[paper]	-
LFM AISTATS-20	[paper]	[code]	RetMol	[paper]	[code]
MolGPT	[paper]	-	Bridge (NeurIPS-2022)	[paper]	-
Bresson et al.	[paper]	-	FLAG (ICLR-23)	[paper]	[code]
LIMO	[paper]	[code]	D3FG NeurIPS-24	[paper]	[code]

Conformation design

Methods	Paper	Code	Methods	Paper	Code
ConfGF (ICML-21)	[paper]	[code]	DGSM (NeurIPS-21)	[paper]	-
GeoDiff (ICLR-22)	[paper]	[code]	ColfNet (ICML-22)	[paper]	-
Torsion Diffusion (NeurIPS-22)	[paper]	[code]	DiffMD (AAAI-23)	[paper]	-
RINGER	[paper]	-	-	-	-

De novo ligand design

Methods	Paper	Code	Methods	Paper	Code
DiffBP	[paper]	-	DiffSBDD	[paper]	-
TargetDiff (ICLR-23)	[paper]	-	PMDM	[paper]	-
D3FG	[paper]	-	-	-	-

Ligand docking

Methods	Paper	Code
DIFFDOCK (NeurIPS-22)	[paper]	-
EDM-Dock (JCIM)	[paper]	[code]
DPL	[paper]	[code]
NeuralPLexer	[paper]	-
E3BIND (ICLR-23)	[paper]	-

Protein design

Methods	Paper	Code	Methods	Paper	Code
DiffAb (NeurIPS-22)	[paper]	-	Anand	[paper]	-
PROTSEED (ICLR-23)	[paper]	-	ProteinSGM	[paper]	[code]
SMCDiff (ICLR-23)	[paper]	[code]	GraDe-IF	[paper]	-
EigenFold	[paper]	[code]	-	-	-
CGM (NeurIPS-19)	[paper]	[code]	SeqDesign	[paper]	[code]
Fold2Seq (ICML-21)	[paper]	[code]	EvoDiff	[paper]	-
GVP （ICLR-21）	[paper]	[code]	ESM (NeurIPS-21)	[paper]	-

Motion generation

Motion synthesis

Methods	Paper	Code	Homepage	Methods	Paper	Code	Homepage
MotionDiffuse	[paper]	[code]	[homepage]	Modiff	[paper]	-	-
Ren et al. (ICASSP-23)	[paper]	-	-	FLAME (AAAI-23)	[paper]	-	-
MoFusion (CVPR-23)	[paper]	-	[homepage]	MDM (ICLR-23)	[paper]	[code]	[homepage]
MLD (CVPR-23)	[paper]	[code]	[homepage]	PriorMDM	[paper]	[code]	[homepage]
Alexanderson et al. (ACM Trans. Graph.)	[paper]	-	-	EDGE (CVPR-23)	[paper]	[code]	[homepage]
SceneDiffuser	[paper]	[code]	[homepage]	MoDi (CVPR-23)	[paper]	[code]	[homepage]
BiGraphDiff	[paper]	-	-	DiffuPose	[paper]	-	-

Motion prediction

Methods	Paper	Code	Homepage
Ahn et al. (ICRA-23)	[paper]	[code]	[homepage]
HumanMAC	[paper]	[code]	[homepage]
TCD (ICRA-23)	[paper]	[code]	-
DiffMotion	[paper]	-	-

Others

Methods	Paper	Code	Homepage	Methods	Paper	Code	Homepage
EDP-GNN	[paper]	-	-	GSDM	[paper]	-	-
NVDiff	[paper]	-	-	DPM-GSP	[paper]	-	-
DiffSTG	[paper]	-	-	DIFUSCO	[paper]	-	-
GraphGDP	[paper]	[code]	-	HouseDiffusion (CVPR-23)	[paper]	[code]	[homepage]
NAP	[paper]	-	[homepage]	EDGE	[paper]	-	-
DruM	[paper]	-	-	DDM	[Paper]	-	-
DiffusionNAG	[paper]	-	-	TSDiff	[paper]	-	-
GraphArm	[paper]	-	-	HGDM	[paper]	[code]	-
Lee et al.	[paper]	-	-	SaGess	[paper]	-	-
SLD	[paepr]	-	-	Diff-POI	[paper]	-	-
Brain Diffuser	[paper]	-	-	Lu et al.	[paper]	-	-
GBD	[paper]	[code]	-	DiffGraph	[paper]	[code]	-
ProGDM	[paper]	[code]	-	R-ode	[paper]	-	-

Datasets

Dataset	Dimensionality	Category	No.of Graphs (G)	No. of Nodes (N)
Community-small	2D	Social	100	11 < N < 20
Ego-small	2D	Social	200	3 < N < 18
Grid	2D	Grid	100	N <= 400
QM9	3D	Bioinformatics/Molecular	130,831	3 < N < 29
ZINC250K	3D	Bioinformatics/Molecular	249,456	6 < N < 38
Enzymes	3D	Bioinformatics/Protein	600	9 < N < 125
SBM-27	2D	Social	200	24 < N < 27
Planar-60	2D	Social	200	N = 60
AIDS	2D	Bioinformatics/Molecular	2000	-
Synthie	2D	Social	300	N = 100
Proteins	3D	Bioinformatics/Protein	1113	N = 39.1

Molecule generation

Methods	Paper	Source	Methods	Paper	Source
Zinc	[paper]	[source]	GEOM-QM9	[paper]	[source]
GEOM-Drugs	[paper]	[source]	CrossDocked2020	[paper]	[source]
BioLiP	[paper]	[source]	PDBBind	[paper]	[source]
SAbDab	[paper]	[source]	-	-	-

Motion generation

Methods	Paper	Source	Methods	Paper	Source
Human3.6M	[paper]	[source]	HumanEva-I	[paper]	[source]
HumanAct12	[paper]	[source]	HumanML3D	[paper]	[source]
KIT	[paper]	[source]	BABEL	[paper]	[source]
UESTC	[paper]	[source]	3DPW	[paper]	[source]
NTU RGB+D	[paper]	[source]	AIST++	[paper]	[source]
TSG	[paper]	[source]	ZeroEGGS	[paper]	[source]

Other surveys

Paper	Url	Source
Diffusion-based Graph Generative Methods	[paper]	[source]
Diffusion Models: A Comprehensive Survey of Methods and Applications	[paper]	[source]
A Survey on Generative Diffusion Model	[paper]	[source]
Generative Diffusion Models on Graphs: Methods and Applications	[paper]	-
A Survey on Graph Diffusion Models: Generative AI in Science for Molecule, Protein and Material	[paper]	-
Graph-based Molecular Representation Learning	[paper]	-
Generative Models as an Emerging Paradigm in the Chemical Sciences	[paper]	-
A Survey on Deep Graph Generation: Methods and Applications (LoG-22)	[paper]	-
A Survey on Temporal Graph Representation Learning and Generative Modeling	[paper]	-
Human motion modeling with deep learning: A survey	[paper]	-
MolGenSurvey: A Systematic Survey in Machine Learning Models for Molecule Design	[paper]	-

Acknowledgement ❤️

Thanks to Diffusion-based-Graph-Generative-Methods.

Star History