The overview of the Contrastive Learning based Multi-Task Graph Neural Network framework (CLMGraph) is shown in Figure 1. CLMGraph consists of an input module, a feature extraction module, and an ADMET predictor module.

Convert compound $x$ into a molecular graph: $G_x=(N_x,E_x)$ , where $Nx$ represents the node matrix and $Ex$ represents the edge feature matrix.

Update the node and edge information of the molecular graph using a Graph Neural Network (GNN): ${G_x}^{\text{'}}=({N_x}^{\text{'}},{E_x}^{\text{'}})$ Optimize the GNN network by maximizing the mutual information between pairs of augmented molecular graphs using a self-supervised pre-training strategy based on contrastive learning. Use NT-Xent as the objective function to train the model:
$${\mathrm{Loss}}_{i,j}=-\log \frac{\exp \left(\mathrm{sim}\right(z_i,z_{i\text{'}})/\tau )}{{\sum }_N^{\mathrm{k=1}}\exp \left(\mathrm{sim}\right(z_i,z_k)/\tau )}$$ where $z_i$ $z_{i\text{'}}$ , and $z_k$ are the feature vectors extracted by GNN from positive pairs and negative samples, $N$ is the batch size, $\tau$ is the temperature parameter, and $\mathrm{sim}(z_i,z_{i\text{'}})$ represents cosine similarity.

Jointly train the pre-trained model on 108 ADMET prediction tasks to leverage and share relevant information across similar tasks. Optimize the ADMET prediction module using a related task loss (related task loss):
$${\mathrm{Loss}}_t={\sum }_{\text{N}}^{\text{k=1}}{\mathrm{Loss}}_n$$ Overall, the CLMGraph framework combines contrastive learning and multi-task learning to extract features from compound structures using self-supervised pre-training and joint training, leading to accurate predictions on ADMET tasks. This approach improves model performance and leverages information sharing among related tasks to accelerate learning.

Properities	Total (positive/Negative)	Training set (positive/Negative)	validation set (positive/Negative)	test set (positive/Negative)
logS	1514	1080	275	159
logP	13076	9392	2368	1316
pKa	5149	3723	936	490
Acidic pKa	2719	1965	491	263
Basic pKa	2725	1954	514	257
Caco-2 classify	962/749	679/543	183/130	100/76
Caco-2	1750	1255	316	179
HIA	65/437	51/301	12/91	2/45
MDCK	217/205	151/143	43/39	23/23
F50%	854/1086	625/768	154/207	75/111
F30%	598/1342	433/960	115/246	50/136
F20%	445/1495	323/1070	88/273	34/152
BBB	451/1976	323/1444	74/338	54/194
OATP1B1 inhibitor	176/1276	131/914	36/211	9/151
OATP1B3 inhibitor	118/1348	84/971	27/221	7/156
OATP2B1 inhibitor	132/38	92/24	34/11	6/3
OCT1 inhibitor	84/34	70/21	9/11	5/2
OCT2 inhibitor	522/150	367/115	99/21	56/14
BCRP inhibitor	483/423	341/306	87/71	55/46
BSEP inhibitor	212/258	160/178	31/47	21/33
MATE1 inhibitor	557/59	400/43	107/7	50/9
Pgp inhibitor	651/1118	485/842	103/178	63/98
Pgp substrate	797/669	569/490	149/116	79/63
PPB	3804	2716	668	420
VDss	1166	845	207	114
CYP1A2 inhibitor	9037/8095	6564/5784	1568/1506	905/805
CYP3A4 inhibitor	13790/5734	9932/4084	2491/1025	1367/625
CYP2B6 inhibitor	306/182	215/135	58/28	33/19
CYP2C9 inhibitor	11516/6277	8338/4507	2074/1099	1104/671
CYP2C19 inhibitor	10087/7974	7301/5719	1799/1433	987/822
CYP2D6 inhibitor	14264/3773	10269/2698	2577/670	1418/405
CYP1A2 substrate	1055/457	746/341	198/71	111/45
CYP3A4 substrate	1747/1330	1257/966	314/240	176/124
CYP2B6 substrate	248/229	183/178	40/34	25/17
CYP2C9 substrate	1486/511	1071/361	271/97	144/53
CYP2C19 substrate	1038/398	743/284	200/76	95/38
CYP2D6 substrate	1519/615	1104/452	279/98	136/65
HLM	3871/2126	2800/1525	689/407	382/194
RLM	2270/3294	1613/2361	419/600	238/333
UGT substrate	537/597	390/441	95/95	52/61
CLp	565/521	409/380	94/98	62/43
CLr	204/280	152/207	34/43	18/30
T1/2	1155	837	208	110
MRT	1165	844	209	112
Neurotoxicity	419	303	67	49
DILI	901/1255	672/903	145/229	84/123
hERG 1uM	3696/797	2633/599	718/135	345/63
hERG 10uM	1826/2667	1285/1947	360/493	181/227
hERG 30uM	912/3581	630/2602	183/670	99/309
hERG 1-10uM	1826/797	1285/599	360/135	181/63
hERG 10-30uM	912/2667	630/1947	183/493	99/227
Respiratory toxicity	997/1185	697/875	188/202	112/108
Nephrotoxicity	350/301	262/215	62/55	26/31
Eye corrosion	1194/855	849/613	218/141	127/101
Eye irritation	1137/3761	813/2739	208/680	116/342
Skin corrosion	1187/698	854/469	199/136	134/93
Skin irritation	1187/1256	854/913	199/218	134/125
Skin sensitisation	565/354	388/250	109/66	68/38
Acute dermal toxicity	664/784	453/561	138/146	73/77
Ames mutagenesis	3601/4372	2561/3197	672/762	368/413
Mouse carcinogenicity classify	386/371	267/263	76/75	43/33
Mouse carcinogenicity	376	265	79	32
Rat carcinogenicity classify	474/500	338/349	92/101	44/50
Rat carcinogenicity	498	345	105	48
Rodents carcinogenicity	394/498	279/359	73/92	42/47
Micronucleus	364/225	267/157	58/48	39/20
Reproductive toxicity	917/783	653/553	175/148	89/82
Mitochondrial toxicity	1504/1361	1101/981	253/249	150/131
Hemolytic toxicity	496/387	348/271	97/74	51/42
Repeated dose toxicity	674/1145	474/822	135/203	65/120
Acute oral toxicity classify	1343/2581	936/1908	242/417	165/256
Acute oral toxicity	13493	9773	2332	1388
FDAMDD classify	505/428	380/315	86/72	39/41
FDAMDD	1074	801	176	97
AR	5745/524	4143/391	1030/87	572/46
ER	6220/391	4484/293	1113/66	623/32
AR-LBD	5818/542	4206/394	1037/96	575/52
ER-LBD	5625/663	4047/502	1021/104	557/57
Aromatase	4977/655	3593/473	887/117	497/65
AhR	5205/707	3782/509	916/138	507/60
ARE	4256/814	3030/606	791/145	435/63
ATAD5	6162/328	4457/254	1094/47	611/27
HSE	5295/318	3807/238	954/51	534/29
p53	5772/404	4170/297	1038/74	564/33
PPARγ	5949/471	4316/335	1046/95	587/41
MMP	4214/797	3026/586	769/135	419/76
TR	4578/331	3272/249	847/52	459/30
GR	4946/549	3555/390	893/110	498/49
Honey bee toxicity	462/158	333/117	83/23	46/18
Avain toxicity (Colinus virginanus)	395/84	282/71	71/9	42/4
Avain toxicity (Anas platyrhynchos)	416/82	310/62	65/14	41/6
Aquatic toxicity (P. subcapitata)	385/464	264/327	74/93	47/44
Aquatic toxicity (Crustaceans)	326/369	236/279	53/54	37/36
Aquatic toxicity (D. magna)	302/287	221/216	45/41	36/30
Aquatic toxicity (Fish)	597/831	436/628	102/125	59/78
Aquatic toxicity (Fathead minnow)	437/424	312/322	79/56	46/46
Aquatic toxicity (Bluegill sunfish)	264/504	195/382	42/77	27/45
Aquatic toxicity (Rainbow trout)	217/553	156/422	41/81	20/50
Aquatic toxicity (Sheepshead minnow)	102/228	84/168	11/38	7/22
Aquatic toxicity (T. pyriformis, classify)	337/965	250/735	51/150	36/80
Aquatic toxicity (T. pyriformis)	1310	991	202	117
BCF classify	488/156	346/107	91/30	51/19
BCF	463	331	79	53
Biodegradability	906/550	644/401	158/94	104/55
Photoinduced toxicity	659/627	473/458	118/115	68/54
Phototoxicity/Photoirritation	221/146	150/108	43/26	28/12
Photoallergy	660/462	475/333	117/85	68/44

Properities	Test set				Validation set
	PCC	R2	MAE	RMSE	PCC	R2	MAE	RMSE
logS	0.961	0.924	0.459	0.603	0.956	0.915	0.499	0.69
logP	0.968	0.936	0.346	0.495	0.967	0.935	0.363	0.507
pKa	0.803	0.645	1.054	2.014	0.869	0.755	0.924	1.63
Acidic pKa	0.883	0.78	0.918	1.626	0.903	0.815	0.829	1.449
Basic pKa	0.781	0.61	1.046	2.183	0.893	0.797	0.85	1.565
Caco-2	0.857	0.735	0.306	0.402	0.818	0.669	0.33	0.419
PPB	0.814	0.662	0.108	0.154	0.846	0.716	0.107	0.151
VDss	0.778	0.605	0.322	0.447	0.819	0.671	0.309	0.422
T1/2	0.506	0.256	0.43	0.593	0.599	0.358	0.379	0.544
MRT	0.523	0.273	0.449	0.615	0.603	0.364	0.383	0.542
Neurotoxicity	0.817	0.668	0.238	0.312	0.752	0.566	0.199	0.293
Mouse carcinogenicity	0.68	0.462	0.729	0.922	0.72	0.518	0.851	1.085
Rat carcinogenicity	0.69	0.476	0.864	1.071	0.61	0.372	1.095	1.334
Acute oral toxicity	0.77	0.592	0.378	0.534	0.765	0.585	0.387	0.54
FDAMDD	0.585	0.342	0.738	1.025	0.755	0.57	0.525	0.693
Aquatic toxicity (T. pyriformis)	0.937	0.879	0.283	0.39	0.945	0.892	0.274	0.35
BCF	0.858	0.737	0.505	0.645	0.853	0.727	0.511	0.684

Properities	Test set				Validation set
	AUC	Precision	Recall	Specificity	AUC	Precision	Recall	Specificity
Caco-2 classify	0.95	0.846	0.898	0.878	0.905	0.795	0.816	0.84
HIA	0.977	0.961	0.936	0.824	1.0	0.978	1.0	0.5
MDCK	0.838	0.862	0.641	0.879	0.786	0.773	0.739	0.783
F50%	0.78	0.738	0.756	0.654	0.808	0.771	0.73	0.68
F30%	0.792	0.802	0.892	0.514	0.868	0.881	0.875	0.68
F20%	0.763	0.826	0.937	0.354	0.879	0.92	0.908	0.647
BBB	0.927	0.932	0.983	0.69	0.946	0.917	0.974	0.685
OATP1B1 inhibitor	0.823	0.939	0.943	0.462	0.709	0.967	0.98	0.444
OATP1B3 inhibitor	0.891	0.947	0.983	0.235	0.889	0.963	1.0	0.143
OATP2B1 inhibitor	0.851	0.333	0.143	0.909	1.0	0.0	0.0	1.0
OCT1 inhibitor	0.889	0.6	0.6	0.889	1.0	1.0	0.5	1.0
OCT2 inhibitor	0.84	0.643	0.29	0.94	0.839	0.625	0.357	0.946
BCRP inhibitor	0.886	0.794	0.89	0.761	0.861	0.724	0.913	0.709
BSEP inhibitor	0.883	0.776	0.865	0.675	0.938	0.838	0.939	0.714
MATE1 inhibitor	0.851	0.0	0.0	1.0	0.733	0.0	0.0	1.0
Pgp inhibitor	0.924	0.837	0.938	0.678	0.892	0.836	0.939	0.714
Pgp substrate	0.839	0.75	0.745	0.765	0.826	0.721	0.778	0.759
CYP1A2 inhibitor	0.912	0.809	0.837	0.826	0.922	0.824	0.873	0.834
CYP3A4 inhibitor	0.901	0.701	0.735	0.871	0.897	0.74	0.715	0.885
CYP2B6 inhibitor	0.741	0.571	0.48	0.83	0.817	0.667	0.737	0.788
CYP2C9 inhibitor	0.888	0.732	0.759	0.852	0.895	0.781	0.768	0.87
CYP2C19 inhibitor	0.897	0.796	0.836	0.827	0.909	0.79	0.845	0.813
CYP2D6 inhibitor	0.847	0.67	0.524	0.935	0.872	0.73	0.548	0.942
CYP1A2 substrate	0.923	0.758	0.758	0.878	0.914	0.68	0.756	0.856
CYP3A4 substrate	0.873	0.78	0.786	0.802	0.867	0.721	0.855	0.767
CYP2B6 substrate	0.885	0.78	0.852	0.735	0.856	0.722	0.765	0.8
CYP2C9 substrate	0.857	0.58	0.554	0.861	0.851	0.667	0.566	0.896
CYP2C19 substrate	0.944	0.816	0.827	0.922	0.912	0.727	0.632	0.905
CYP2D6 substrate	0.885	0.726	0.625	0.886	0.92	0.79	0.754	0.904
HLM	0.85	0.743	0.661	0.882	0.822	0.669	0.552	0.861
RLM	0.813	0.763	0.829	0.624	0.828	0.772	0.853	0.647
UGT substrate	0.806	0.768	0.729	0.757	0.716	0.676	0.754	0.577
CLp	0.807	0.713	0.72	0.722	0.792	0.675	0.628	0.79
CLr	0.685	0.592	0.913	0.341	0.517	0.641	0.833	0.222
DILI	0.716	0.742	0.715	0.615	0.707	0.7	0.74	0.536
hERG 1uM	0.868	0.654	0.614	0.92	0.879	0.561	0.587	0.916
hERG 10uM	0.884	0.803	0.894	0.675	0.88	0.761	0.912	0.641
hERG 30uM	0.919	0.906	0.963	0.637	0.908	0.854	0.968	0.485
hERG 1-10uM	0.946	0.806	0.873	0.896	0.96	0.679	0.905	0.851
hERG 10-30uM	0.941	0.92	0.958	0.769	0.938	0.869	0.96	0.667
Respiratory toxicity	0.913	0.889	0.804	0.866	0.904	0.827	0.796	0.839
Nephrotoxicity	0.777	0.719	0.695	0.789	0.744	0.655	0.613	0.615
Eye corrosion	0.985	0.886	0.986	0.92	0.98	0.923	0.95	0.937
Eye irritation	0.951	0.921	0.981	0.736	0.978	0.944	0.991	0.828
Skin corrosion	0.931	0.785	0.848	0.866	0.937	0.865	0.828	0.91
Skin irritation	0.782	0.698	0.77	0.673	0.794	0.683	0.76	0.672
Skin sensitisation	0.915	0.8	0.727	0.881	0.854	0.683	0.737	0.809
Acute dermal toxicity	0.848	0.835	0.722	0.812	0.764	0.689	0.662	0.685
Ames mutagenesis	0.88	0.808	0.837	0.745	0.855	0.773	0.833	0.726
Mouse carcinogenicity classify	0.718	0.644	0.613	0.687	0.704	0.618	0.636	0.698
Rat carcinogenicity classify	0.708	0.615	0.709	0.598	0.779	0.75	0.78	0.705
Rodents carcinogenicity	0.731	0.64	0.679	0.557	0.85	0.776	0.809	0.738
Micronucleus	0.949	0.767	0.868	0.873	0.936	0.85	0.85	0.923
Reproductive toxicity	0.887	0.939	0.655	0.963	0.885	0.9	0.659	0.933
Mitochondrial toxicity	0.908	0.843	0.836	0.853	0.914	0.829	0.817	0.853
Hemolytic toxicity	0.847	0.709	0.847	0.709	0.845	0.721	0.738	0.765
Repeated dose toxicity	0.732	0.741	0.827	0.544	0.736	0.761	0.875	0.492
Acute oral toxicity classify	0.892	0.892	0.859	0.777	0.87	0.828	0.867	0.721
FDAMDD classify	0.829	0.76	0.76	0.791	0.894	0.783	0.878	0.744
AR	0.851	0.649	0.363	0.981	0.934	0.526	0.435	0.969
ER	0.902	0.72	0.462	0.987	0.949	0.586	0.531	0.981
AR-LBD	0.863	0.574	0.39	0.972	0.931	0.591	0.5	0.969
ER-LBD	0.867	0.546	0.44	0.952	0.934	0.558	0.509	0.959
Aromatase	0.822	0.565	0.416	0.955	0.85	0.481	0.4	0.944
AhR	0.864	0.557	0.412	0.955	0.869	0.617	0.483	0.964
ARE	0.856	0.604	0.384	0.946	0.884	0.655	0.603	0.954
ATAD5	0.835	0.765	0.191	0.996	0.917	0.6	0.222	0.993
HSE	0.852	0.625	0.185	0.994	0.831	0.5	0.172	0.991
p53	0.866	0.455	0.205	0.983	0.884	0.5	0.242	0.986
PPARγ	0.922	0.274	0.586		0.939	0.296	0.615
MMP	0.93	0.755	0.669	0.955	0.949	0.75	0.789	0.952
TR	0.865	0.636	0.233	0.99	0.857	0.571	0.133	0.993
GR	0.906	0.812	0.544	0.985	0.912	0.812	0.531	0.988
Honey bee toxicity	0.799	0.579	0.367	0.907	0.955	0.923	0.667	0.978
Avain toxicity (Colinus virginanus)	0.764	0.636	0.368	0.953	1.0	1.0	1.0	1.0
Avain toxicity (Anas platyrhynchos)	0.821	0.727	0.421	0.966	0.907	0.8	0.667	0.976
Aquatic toxicity (P. subcapitata)	0.895	0.821	0.897	0.729	0.811	0.667	0.864	0.596
Aquatic toxicity (Crustaceans)	0.888	0.768	0.828	0.765	0.872	0.829	0.806	0.838
Aquatic toxicity (D. magna)	0.895	0.741	0.833	0.788	0.856	0.815	0.733	0.861
Aquatic toxicity (Fish)	0.911	0.826	0.902	0.721	0.91	0.829	0.872	0.763
Aquatic toxicity (Fathead minnow)	0.911	0.843	0.843	0.817	0.891	0.844	0.826	0.848
Aquatic toxicity (Bluegill sunfish)	0.92	0.848	0.918	0.709	0.942	0.878	0.956	0.778
Aquatic toxicity (Rainbow trout)	0.893	0.852	0.945	0.647	0.916	0.9	0.9	0.75
Aquatic toxicity (Sheepshead minnow)	0.938	0.818	0.957	0.643	0.89	0.955	0.955	0.857
Aquatic toxicity (T. pyriformis, classify)	0.974	0.931	0.964	0.821	0.949	0.923	0.9	0.833
BCF classify	0.958	0.583	0.875	0.851	0.943	0.739	0.895	0.882
Biodegradability	0.839	0.628	0.772	0.753	0.872	0.643	0.818	0.76
Photoinduced toxicity	0.712	0.659	0.68	0.656	0.743	0.633	0.704	0.676
Phototoxicity/Photoirritation	0.719	0.609	0.467	0.8	0.747	0.438	0.583	0.679
Photoallergy	0.703	0.6	0.6	0.73	0.744	0.667	0.636	0.794

admetSAR3.0 was designed with HTML, CSS, and JavaScript for the frontend, utilizing jQuery for page control and event handling. It incorporates JSME as a molecular editor and employs ApacheECharts for data visualization. The core logic of the website is implemented in PHP, with mariaDB chosen as the data management system. The prediction models were built with Python, and the modeling process was supported by chemical informatics packages such as PyTorch, DGL, DGLlife, and RDKit. The server has been successfully tested on Google Chrome, Microsoft Edge, Mozilla Firefox, and Apple Safari.

Software and tool	Version
PHP	5.4.16
MySQL	5.5.64
Anaconda	4.12.0
NVIDIA CUDA Toolkit	11.5
Python	3.9.18
RDKit	2022.03.2
dgl	0.9.1
dgllife	0.3.2
pytorch	1.11.0
torchvision	0.12.0
jQuery	2.1.1
Bootstrap	3.3.7
JSME	2022-06-10
Apache ECharts	5.4.0

OS	Version	Chrome	Edge	Firefox	Safari
Linux	Ubuntu 18.04.6 LTS	79.0.3945.88	-	74.0.1	-
MacOS	Catalina 10.15.7	75.0.3770.80	-	71.0	13.1
Windows	10	75.0.3770.80	80.0.361.109	71.0	-