Table 3.

Multi-classification performance of functional phenotypic datasets using fine-turned models

	Model	Accuracy	F1	Loss	Precision	Recall	Auroc_weighted	Pr_auc	MCC	Time(h)
Dataset_HERV	A	0.8880 ± 0.0125	0.8895 ± 0.0077	0.3352 ± 0.0177	0.8989 ± 0.0045	0.8880 ± 0.0125	0.90625 ± 0.0047	0.8455 ± 0.0063	0.7967 ± 0.0197	14.6616 ± 0.4380
	B	0.8954 ± 0.0024	0.8956 ± 0.0017	0.3061 ± 0.0087	0.8985 ± 0.0012	0.8954 ± 0.0024	0.9122 ± 0.0010	0.8515 ± 0.0014	0.8078 ± 0.0040	18.8373 ± 1.7596
Dataset_Immuno	A	0.8096 ± 0.0029	0.8091 ± 0.0033	0.3432 ± 0.0025	0.8164 ± 0.0033	0.8096 ± 0.0029	0.8197 ± 0.0020	0.7487 ± 0.0017	0.6549 ± 0.0023	0.6727 ± 0.0035
	B	0.8057 ± 0.0022	0.8049 ± 0.0029	0.3396 ± 0.0040	0.8180 ± 0.0063	0.8057 ± 0.0022	0.8180 ± 0.0012	0.7471 ± 0.0017	0.6534 ± 0.0056	1.9903 ± 0.0047
Dataset_Regulatory	A	0.5937 ± 0.0033	0.6055 ± 0.0029	0.9905 ± 0.0073	0.6491 ± 0.0028	0.5937 ± 0.0033	0.7366 ± 0.0017	0.5107 ± 0.0024	0.4634 ± 0.0038	22.4265 ± 1.1836
	B	0.6263 ± 0.012	0.6355 ± 0.0099	0.9275 ± 0.0309	0.6662 ± 0.0008	0.6263 ± 0.0128	0.7536 ± 0.0038	0.5324 ± 0.0047	0.4966 ± 0.0098	31.2417 ± 2.6290
Dataset_Diseases_GWAS	A	0.7563 ± 0.0062	0.7579 ± 0.0044	0.5693 ± 0.0049	0.7698 ± 0.0029	0.7563 ± 0.0062	0.8193 ± 0.0005	0.6599 ± 0.0018	0.6355 ± 0.0049	7.7064 ± 0.2058
	B	0.7511 ± 0.0015	0.7585 ± 0.0012	0.5747 ± 0.0036	0.7761 ± 0.0006	0.7511 ± 0.0015	0.8293 ± 0.0003	0.6683 ± 0.0006	0.6376 ± 0.0011	22.5575 ± 0.6066
Dataset_Highly_Specifically_Gene	A	0.6002 ± 0.0048	0.6014 ± 0.0034	0.8146 ± 0.0092	0.6029 ± 0.0031	0.6002 ± 0.0048	0.6312 ± 0.0029	0.5498 ± 0.0024	0.2551 ± 0.0059	0.2179 ± 0.0122
	B	0.5652 ± 0.0091	0.5673 ± 0.0129	0.8564 ± 0.0164	0.5725 ± 0.0166	0.5652 ± 0.0091	0.6003 ± 0.0152	0.5270 ± 0.0107	0.1947 ± 0.0274	0.1933 ± 0.0159
Dataset_Random	A	0.2864 ± 0.012	0.2793 ± 0.0025	1.3934 ± 0.0011	0.2881 ± 0.0053	0.2864 ± 0.0126	0.4943 ± 0.0038	0.2935 ± 0.0014	-0.0109 ± 0.0074	0.7665 ± 0.0030
	B	0.3321 ± 0.0095	0.2984 ± 0.0071	1.3955 ± 0.0029	0.2992 ± 0.0018	0.3321 ± 0.0095	0.5026 ± 0.0009	0.2969 ± 0.0004	0.0052 ± 0.0018	2.4553 ± 0.0020

Note The A and B represent the DNA_bert_6 model and the human_gpt2-v1 model, respectively