Table 1 The AUC results for binary sequence classification tasks on human genome

Data	DNABERT-2	NT-v2	HyenaDNA	Caduceus-Ph	GROVER
DNase I Hypersensitive	0.8666	0.8524	0.8295	0.8799	0.857
Human TFBS 1	0.8382	0.8315	0.8301	0.8796	0.8618
Human TFBS 2	0.821	0.809	0.8205	0.8687	0.8495
Human TFBS 3	0.7896	0.7974	0.7875	0.8249	0.8158
Human TFBS 4	0.726	0.7103	0.7149	0.7725	0.763
Human TFBS 5	0.9204	0.9149	0.9159	0.9294	0.931
Promoter GM12878	0.9856	0.9835	0.976	0.9865	0.9839
Promoter HUVEC	0.9903	0.987	0.9817	0.9896	0.9885
Promoter Hela-S3	0.9886	0.9838	0.981	0.9871	0.9857
Promoter NHEK	0.9501	0.9323	0.9271	0.9567	0.9507
Acceptor	0.8969	0.7928	0.7946	0.8449	0.8041
Coding	0.9438	0.9289	0.9406	0.9735	0.9594
Donor	0.9056	0.8198	0.8128	0.8535	0.819
Enhancer	0.8717	0.8674	0.8339	0.8384	0.8554
Enhancer Cohn	0.8223	0.7894	0.7754	0.821	0.8161
Enhancer Ensembl	0.9369	0.9389	0.9356	0.9431	0.9382
Open chromatin region	0.7253	0.7183	0.7191	0.765	0.7455
Promoter All 300 bps	0.9426	0.9445	0.9394	0.9519	0.9402
Promoter All 70 bps	0.8311	0.8527	0.832	0.8748	0.8506
Promoter NonTATA 251 bps	0.9297	0.8905	0.928	0.9426	0.9395
Promoter NonTATA 300 bps	0.9765	0.9758	0.9662	0.9834	0.9728
Promoter NonTATA 70 bps	0.8531	0.8729	0.8516	0.8961	0.8704
Promoter TATA 300 bps	0.7646	0.7791	0.8077	0.76	0.78
Promoter TATA 70 bps	0.7781	0.7947	0.7827	0.8103	0.796

The tasks include promoter region identification (across multiple datasets), coding region detection, splice site donor and acceptor identification, enhancer identification (across multiple datasets), transcription factor binding site identification (across multiple datasets), and open chromatin region identification (across multiple datasets). Using mean token pooling method. Bolded: higher than at least two other AUCs, p < 0.01. P-values are calculated using one-sided DeLong’s test.

Quick links

Search