ML4Porous

Summarize the machine learning research on porous materials such as COF/MOF/Zeolite so far.

Band gap models

unit: eV

DOI	Model	${R^{2}}$	MAE	RMSE	SRCC	Year	Database (# of Total Points)	Train: Val: Test
10.1016/j.matt.2021.02.015	CGCNN	0.876	0.274	-	0.932	2021	QMOF-PBE-D3(BJ) (14,482)	0.8 : 0.1 : 0.1
	KRR (SOAP)	0.822	0.357	-	0.910	2021	QMOF-PBE-D3(BJ) (14,482)	0.8 : - : 0.2
	KRR (Orbital field matrix)	0.763	0.417	-	0.863	2021	QMOF-PBE-D3(BJ) (14,482)	0.8 : - : 0.2
	KRR (Stoichiometric-120)	0.750	0.433	-	0.847	2021	QMOF-PBE-D3(BJ) (14,482)	0.8 : - : 0.2
	KRR (Stoichiometric-45)	0.743	0.437	-	0.842	2021	QMOF-PBE-D3(BJ) (14,482)	0.8 : - : 0.2
	KRR (Sine Coulomb matrix)	0.643	0.529	-	0.787	2021	QMOF-PBE-D3(BJ) (14,482)	0.8 : - : 0.2
	KRR (Constant mean model)	-	0.973	-	-	2021	QMOF-PBE-D3(BJ) (14,482)	0.8 : - : 0.2
10.1107/S1600576722009797	CNN from PXRD	-	0.492	0.674	-	2022	QMOF-PBE-D3(BJ) (16,029)	0.8 : - : 0.2
10.1021/acsami.3c10323	PMTransformer	-	0.216	-	-	2023	QMOF-PBE-D3(BJ) (20,000)	0.8 : 0.1 : 0.1
10.1038/s42256-023-00628-2	MOFTransformer	-	0.224	-	-	2023	QMOF-PBE-D3(BJ) (20,000)	0.8 : 0.1 : 0.1
10.1021/jacs.2c11420	MOFormer	-	0.387	-	-	2023	QMOF-PBE-D3(BJ) (7,466)	0.8 : - : 0.2
10.1021/acs.jpclett.3c00187	CGCNN	0.709	0.393	0.585	-	2023	QMOF-HSE06 (10,811)	0.8 : 0.1 : 0.1
	MEGNet	0.767	0.368	0.523	-	2023	QMOF-HSE06 (10,811)	0.8 : 0.1 : 0.1
	MEGNet-tuned	0.787	0.332	0.500	-	2023	QMOF-HSE06 (10,811)	0.8 : 0.1 : 0.1
10.1002/adfm.202313596	CGCNN	0.820	0.330	-	-	2023	QMOF-HSE06 (10,811)	0.8 : 0.1 : 0.1
10.26434/chemrxiv-2024-0mkw8	RR	0.169	0.091	-	-	2023	EC-MOF-PBE-D3-U (1,063)	0.9 : - : 0.1
	LIR	0.000	0.099	-	-	2023	EC-MOF-PBE-D3-U (1,063)	0.9 : - : 0.1
	SGD	-0.001	0.099	-	-	2023	EC-MOF-PBE-D3-U (1,063)	0.9 : - : 0.1
	PAR	0.178	0.091	-	-	2023	EC-MOF-PBE-D3-U (1,063)	0.9 : - : 0.1
	SVR	0.355	0.080	-	-	2023	EC-MOF-PBE-D3-U (1,063)	0.9 : - : 0.1
	KNR	0.315	0.080	-	-	2023	EC-MOF-PBE-D3-U (1,063)	0.9 : - : 0.1
	NNR	0.180	0.086	-	-	2023	EC-MOF-PBE-D3-U (1,063)	0.9 : - : 0.1
	ETR	0.542	0.063	-	-	2023	EC-MOF-PBE-D3-U (1,063)	0.9 : - : 0.1
	GBR	0.436	0.075	-	-	2023	EC-MOF-PBE-D3-U (1,063)	0.9 : - : 0.1
	BR	0.475	0.067	-	-	2023	EC-MOF-PBE-D3-U (1,063)	0.9 : - : 0.1
	ABR	0.391	0.083	-	-	2023	EC-MOF-PBE-D3-U (1,063)	0.9 : - : 0.1
	RFR	0.424	0.066	-	-	2023	EC-MOF-PBE-D3-U (1,063)	0.9 : - : 0.1
AI2ASE 2025	M-MOFormer	-	0.359	-	-	2025	QMOF-PBE-D3(BJ) (unknown)	0.7 : 0.15 : 0.15
10.1039/D3RA02142D	CGCNN		0.149	0.208	-	2023	QMOF-HSE06	0.8 : 0.2
	GCN		0.146	0.277	-	2023	QMOF & PMOF - PBE	0.8 : 0.2 : -
	MEGNet		0.154	0.267	-	2023	QMOF & PMOF - PBE	0.8 : 0.2 : -
	SchNet		0.241	0.368	-	2023	QMOF & PMOF - PBE	0.8 : 0.2 : -

Partial atomic charge models

unit: e

Reference	Model	${R^{2}}$	MAE	RMSE	Target	Framework	Year	Database (# of Total Points)	Train: Val: Test
10.1021/acs.jpcc.0c04903	MPNN	-	0.0250	-	DDEC	MOF	2020	CoRE MOF 2014 DDEC (2,266)	0.7 : 0.1 : 0.2
10.1021/acs.chemmater.0c02468	GBDT	-	0.0096	0.0176	DDEC	MOF	2020	CoRE MOF 2014 DDEC (2,932)	0.9 : - : 0.1
		-	0.0500	-	DDEC	COF	2020	CURATED COF (460)	0.9 : - : 0.1
10.1021/acs.jctc.0c01229	PACMOF-v1	0.9952	0.0192	0.0337	DDEC6	MOF	2021	CoRE MOF 2019 (2,974)	0.8 : 0.2
		0.9241	0.0570	-	DDEC	porous molecular crystals	2021	CoRE MOF 2019 (2,974)	0.8 : 0.2
		0.9969	0.0100	-	CM5	MOF	2021	CoRE MOF 2019 (2,974)	0.8 : 0.2
		0.9175	0.0360	-	CM5	porous molecular crystals	2021	CoRE MOF 2019 (2,974)	0.8 : 0.2
10.1038/s41524-024-01277-8	AL-GNN	-	0.0290	-	DDEC6	MOF	2024	QMOF (11,173)	0.8 : 0.08 : 0.12
		-	0.0368	-	DDEC6	Zeolite	2024	ARC-MOF (without DB12 and DB1)	benchmark
		-	0.0239	-	DDEC6	MOF	2024	ARC-MOF (without DB12 and DB1)	benchmark
10.1021/acs.jctc.4c00434	PACMAN	0.9994	0.0055	0.0094	DDEC6	MOF	2024	QMOF (16,779)	0.8 : 0.1 : 0.1
		0.9996	0.0032	0.0051	CM5	MOF	2024	QMOF (16,779)	0.8 : 0.1 : 0.1
		0.9976	0.0224	0.0358	Bader	MOF	2024	QMOF (11,230)	0.8 : 0.1 : 0.1
		0.9722	0.0419	0.0419	REPEAT	MOF	2024	ARC-MOF (40,000)	0.8 : 0.1 : 0.1
		0.9927	0.0095	0.0159	DDEC6	COF	2024	CURATED COF (612)	benchmark
		-	0.0245	-	DDEC6	Zeolite	2024	IZA DB (258)	benchmark
10.1038/s41524-024-01413-4	MEPO-ML	-	0.0250	-	REPEAT	MOF	2024	ARC-MOF (279,632)	0.8 : 0.1 : 0.1
10.1021/acs.jpcc.4c04879	PACMOF-v2	0.9937	0.0229	0.0376	DDEC6	neutral-MOF	2024	CoRE MOF 2019 (3,378) & QMOF (19,961) & CSD MOF (18,928)	0.8 : - : 0.2
		0.9930	0.0240	0.0380	DDEC6	ionic-MOF	2024	CSD MOF (7,598)	0.8 : - : 0.2
		0.9700	0.0200	0.0330	DDEC6	COF	2024	CURATED COF (792)	benchmark
		0.9980	0.0290	0.0550	DDEC6	Zeolite	2024	Zeo-1 DB (220)	benchmark

Large language models

DOI	Model	Purpose	Year
10.1109/BigData55660.2022.10020568	Pre-Trained Language Models	domain PLMs outperform in MOF KG extraction, but gaps remain	2022
10.48550/arXiv.2201.08174	MOF-KG	KGQA4MAT: ChatGPT-enabled MOF knowledge graph querying benchmark	2022
10.1021/jacs.3c05819	CCA	information extraction from publications	2023
10.1021/jacs.3c12086	fine-tuning GPT	GPT-assisted MOF linker generation via mutation	2023
10.1002/anie.202311983	fine-tuning GPT-4	GPT-4 guides MOF synthesis via human-AI iterative feedback	2023
10.1021/acscentsci.3c01087	ChatGPT-Lab	Multi-AI system accelerates MOF/COF synthesis via ChatGPT and Bayesian search	2023
10.1038/s41467-024-48998-4	ChatMOF	property-based structure generation	2024
10.26434/chemrxiv-2024-7kds2	GPT-4 and GPT-4o-based	LLM-powered MOFsyn Agent optimizes synthesis and catalysis	2024
10.1038/s41467-024-45563-x	fine-tuned GPT-3, Llama-2	LLMs enable flexible extraction of structured materials science knowledge	2024
10.1021/acs.jcim.4c00065	Llama2–7B, ChatGLM2–6B, …	benchmark for different LLM models	2024
10.1021/jacs.4c11085	L2M3	prepare a dataset with information from ~40,000 papers	2025
10.1039/D5TA01139F	fine-tuning Gemini-1.5	hydrophobicity of MOFs	2025
10.26434/chemrxiv-2025-8h5q7	MOF-ChemUnity	linking literature-derived insights to crystal structures and computational datasets	2025

Adsorption / separation models

unit: mmol/g, ${cm^{3}}$/g,...

DOI	Model	${R^{2}}$	MAPE (%)	MAE	RMSE	SRCC	Adsorbate	Condition	Adsorbent	Year	Database (# of Total Points)	Train: Val: Test
10.1021/acs.chemmater.5b01475	RF				2.21		Xe/Kr selectivity	20%Xe:80%Kr,1atm,298K	MOF,PPN,Zeolite	2015	Exp MOF (3,640), hMOF (2,750), hPPN (1,000), hZeolite (2,319), IZA (194), hZIF (3,757), hCOF (834)	- : - : -
10.1021/acscombsci.7b00056	RF	0.97,0.92	8.75,9.22				CH4 (cm3/g),(cm3/cm3)	35bar,298K	MOF	2017	hMOF (130,398)	0.08 : - : 0.92
10.1039/C8ME00050F	LASSO	0.96		2.4	3.1		WC of H2	100-2bar,77K	hMOF	2018	hMOF (137,953) & CSD MOF (54,776)	0.08 : - : 0.92
10.1021/acs.chemmater.8b02257	RF	0.905,0.905,0.855,0.870,0.515				0.921,0.950,0.938,0.914,0.751	CO2/N2 selectivity,CO2,CO2/H2 selectivity,WC of CO2,WC of CO2	15%CO2:85%N2@1bar,1bar,20%CO2:80%H2@20bar,20%CO2:80%H2@20bar,15%CO2:85%N2@1bar	MOF	2018	ToBaCCo MOF (400)	0.8 : - : 0.2
10.1021/acs.jpcc.8b09420	ANN			2.93,1.84,1.78,1.59,1.33,1.57,1.38,1.07,0.99,1.06,1.09,1.24,1.14,1.06,0.84,0.83,0.91,0.75			H2	0.1,1,5,35,65,100bar,77,160,295K	MOF	2019	ToBaCCo (105)	0.8 : - : 0.2
10.1021/acs.jctc.9b00940	ANN	0.998,0.998,0.998,0.998,0.998,0.998	4.4,4.6,4.7,4.5,4.3,4.0	0.17,0.24,0.30,0.42,0.42,0.14			Ar,CH4,Kr,Xe,C2H6,N2	1-100bar	MOF	2020	ToBaCCo MOF (2,400)	9 : 1 : 2
10.1021/acs.chemmater.9b05322	LR,LR,RF,LR	0.988,0.985,0.925,0.992					WC of C2H6,WC of C2H4,C2H6/C2H4 selectivity,C2H6/C2H4 selectivity	1-0.1bar,1-0.1bar,1bar,0.1bar,298K	MOF (UiO-66)	2020	CoRE MOF (425)	0.8 : - : 0.2
10.1021/acs.jpcc.9b10766	RF	0.889,0.824,0.857,0.874		8,16.1,18.2,25.5	12.1,23.6,26.4,37.5		CH4,CO2,H2S,H2	5.8bar@298K,1bar@300K,1bar@300K,2bar,77K	MOF	2020	CoRE MOF 2014 DDEC (2,932)	0.34 : - : 0.66
10.1021/jacs.9b11084	RF	0.74,0.76		12.5,11.5	17.1,17.3		CH4,CO2	5.8bar@298K,0.5bar@298K	MOF	2020	hMOF (78,000)	50-10,000 for taining
10.1021/acsomega.1c00100	XGBoost	0.951,0.973			0.055,0.255		Xe,Xe/Kr separation	20%Xe:80%Kr,1bar,298K	MOF	2021	G-MOFs (-)	0.3 : - : 0.7
10.1021/acs.jcim.2c00876	MOFNet			2.355,2.233,2.231,2.206,2.196,2.197,2.224,2.212,2.626,2.559,2.516,2.527,2.394,2.352,2.325,2.303,0.208,0.288,0.339,0.378,0.497,0.595,0.778,0.844			N2,CO2,CH4	0.2/5/10/20/40/60/80/100,10/50/100/500/1000/1500/20000/50000,50/100/150/200/500/1000/5000/10000kPa	MOF	2022	CSD MOF (7,304) (6,997) (8,539)	0.8 : 0.1 : 0.1
10.1021/acsami.2c08977	LASSO,GBR,ETR,GBR	0.99,0.80,0.73,0.84		1.30E-3,5.45E-3,0.212,5.62E-2	1.70E-3,9.23E-3,0.339,0.104		He,H2,CH4,N2	1bar,298K	MOF Membranes	2022	hMOF(677)(2,715)(5,215)(5,224)	0.8 : - : 0.2
10.1021/acs.iecr.3c02211	GBR	0.951,0.989		0.565,0.157	0.782,0.370		Xe,Kr saturation loading	298K	MOF	2023	CoRE MOF 2019 (1,081)	0.8 : - : 0.2
10.1021/acsami.3c10323	PMTransformer			5.963,0.069,0.053,2.126,1.009,2.995,0.461			H2,N2,O2,CH4,CH4,CH4,CH4	100/1/65/5.8/65/1bar	MOF	2023	hMOF by PORMAKE (60,000)	0.8 : 0.1 : 0.1
10.1038/s42256-023-00628-2	MOFTransformer			6.377,0.071,0.051,2.268,0.999,3.187,0.493			H2,N2,O2,CH4,CH4,CH4,CH4	100/1/65/5.8/65/1bar	MOF	2023	hMOF by PORMAKE (60,000)	0.8 : 0.1 : 0.1
10.1021/acsami.3c02130	tree-based	0.712,0.874		0.379,0.0234	0.552,0.0331	0.852,0.937	CO2,N2	1bar,298K	IL/MOF	2023	CSD MOF (941)	0.8 : - : 0.2
10.1021/jacs.2c11420	MOFormer			0.158,0.545,0.982,0.033,0.161,0.384			CO2,CO2,CO2,CH4,CH4,CH4	0.05bar,0.5bar,2.5bar,0.04bar,0.5bar,2.5bar	MOF	2023	hMOF (102,858)	0.7 : 0.15 : 0.15
10.1021/acsami.3c10951	GC-Trans	0.9811		0.1913	0.5319	0.9906	CO2,N2	0.1-30bar,273-313K	MOF	2023	CSD MOF (3,101)	0.8 : 0.1 : 0.1
10.1038/s41467-024-46276-x	Uni-MOF	0.983					CO2,N2,CH4,Kr,Xe	273K,298K,0.01-10Pa	MOF	2024	hMOF_MOFX-DB & CoRE_MOFX-DB (2,477,494 points)	0.8 : 0.1 : 0.1
		0.916					Ar,N2	77K,87K,1Pa-1bar	MOF	2024	hMOF_MOFX-DB & CoRE_MOFX-DB (464,824 points)	0.8 : 0.1 : 0.1
		0.834					CH4,CO2,Ar,Kr,Xe,O2,He	150-300K, 1Pa-3bar	MOF	2024	hMOF_MOFX-DB & CoRE_MOFX-DB (99,200 points)	0.8 : 0.1 : 0.1
10.1021/jacs.3c14610	RF	0.83					C3H8/C3H6 selectivity	50%C3H8:50%C3H6@1bar@298K	MOF	2024	CoRE MOF (4,685)	0.9 : - : 0.1
10.1021/acsami.4c11953	GBR	0.995,0.994		0.073,0.396	0.119,0.642		H2 (wt.%) (g/L)	5,100bar,111,231,296K	Functionalized COF	2024	CURATED COF (807)	0.8 : - : 0.2
10.1002/anie.202500783	XGBoost	0.968,0.984					C2H6/C2H4 selectivity,C2H6	10%C2H6:90%C2H4@1bar	MOF-5 Analogs	2025	CSD MOF (2,824)	0.8 : - : 0.2

Machine learning potential

DOI	Model	Purpose	Year	Dataset
10.1039/D1MA01152A	DNN	CO2,H2O@MOF-74	2022	Energy,PBE-D3
10.1038/s41524-023-00969-x	NequIP	UiO-66(Zr),MIL-53(Al)	2023	QM energy,PBE-D3
10.1021/acs.jctc.3c00495	NequIP	ZIF-8,Mg-MOF-74	2023	0-32 CO2@ZIF-8/Mg-MOF-74,relax,PBE-D3 (BJ)
10.1039/D3NR05966A	DeePMD-kit	2D MOF(NiF2(pyrazine)2)	2024	AIMD,NVT(100,200,300K),10,000*0.5fs,PBE
10.1039/D3SC05612K	DeePMD-kit	H2@Al-soc-MOF-1d	2024	0.31-5.25wt.%H2,AIMD,NVT(10-100K),1,000*0.5fs,PBE-D3
10.1038/s41524-024-01427-y	SNAP	ZIF-8,MOF-5	2024	AIMD,NVT(100-1,000K),2,000*0.5fs,PBE
10.1038/s41524-024-01205-w	SkoltechMLIP	MIL-53,MOF-74,MOF-5,UiO-66	2024	relax,PBE-D3(BJ)
10.1021/acsnano.4c12369	NequIP	Zn-based MOFs	2025	isotopic strain and randomly perturbing atoms for 2,966 MOFs from QMOF(19,705),PBE-D3(BJ)

Other models

DOI	Model	Accuracy	AUC	${R^{2}}$	MAE	SRCC	Purpose	Materials	Year	Database (# of Total Points)	Train: Val: Test
10.1021/acs.jpclett.8b01707	NN	0.67					metallic	MOF	2018	CoRE MOF 2014 DDEC (2,932)	0.7 : - : 0.3
10.1021/acs.jpclett.0c01518	LASSO			0.960			True monolayer area	MOF	2020	CoRE MOF (300)	0.4 : - : 0.6
10.1038/s41557-021-00717-y	GBR	0.98					oxidation states	MOF	2020	COD & CSD MOF	- : - : -
10.1038/s42256-020-00249-z	RF	0.88					Water stability	MOF	2020	Butch et al (207)	0.8 : - : 0.2
10.1021/jacs.1c07217	ANN				44-47		Thermal stability	MOF	2020	CoRE MOF 2019 (3,132 points)	0.8 : - : 0.2
		0.76	0.79				Solvent removal stability	MOF	2020	CoRE MOF 2019 (-)	0.8 : - : 0.2
10.1038/s41563-022-01374-3	GBR				0.02	0.98	Heat capacity	MOF/COF/Zeolite	2022	CoRE MOF 2019, QMOF, IZA, CURATED COF (232)	0.9 : - : 0.1
10.1021/jacs.4c05879	RF	0.75	0.80				Water stability	MOF	2024	WS24 (964)	0.8 : - : 0.2
10.1002/adma.202500943	XGBoost			0.92,0.80,0.86	11.8,7.74,0.17		Gravimetric, volumetric capacitance, logarithmic current desity	MOF	2024	QMOF (20,375)	0.8 : - : 0.2