Benchmarking DNA foundation models for genomic and genetic tasks: Supplementary Data 1-6

What this source contributed

Records in this release that cite this source. A record can cite several sources.

• 48 protocols
• 47 datasets
• 59 evaluations
• 177 results
• 54 evidence claims

Results

Values as the source printed them. Compare results only under compatible protocols, datasets and metrics.

Claims

Statements this source supports, each with where in the source it was checked.

• reported_finding: amp-feng-20261007-protocol-promoter-nontata: Reported finding Baseline CNN trained from scratch on Promoter Arabidopsis NonTATA: Supplementary Data 1 lists the DNA foundation models the CNN significantly outperformed (one-sided DeLong p<0.01): Caduceus-Ph, DNABERT-2, GROVER, HyenaDNA, NT-v2. Supplementary Data 2 lists the models that significantly outperformed the CNN: None. The CNN's AUC is not printed, so it has no result record; the foundation-model AUCs are in Table 2. (Supplementary Data 1 and Supplementary Data 2, row 'Promoter Arabidopsis NonTATA')
• reported_finding: amp-feng-20261007-protocol-promoter-tata: Reported finding Baseline CNN trained from scratch on Promoter Arabidopsis TATA: Supplementary Data 1 lists the DNA foundation models the CNN significantly outperformed (one-sided DeLong p<0.01): Caduceus-Ph, DNABERT-2, GROVER, NT-v2. Supplementary Data 2 lists the models that significantly outperformed the CNN: None. The CNN's AUC is not printed, so it has no result record; the foundation-model AUCs are in Table 2. (Supplementary Data 1 and Supplementary Data 2, row 'Promoter Arabidopsis TATA')
• reported_finding: amp-feng-20261007-protocol-splice-acceptor: Reported finding Baseline CNN trained from scratch on Acceptor: Supplementary Data 1 lists the DNA foundation models the CNN significantly outperformed (one-sided DeLong p<0.01): Caduceus-Ph, GROVER, HyenaDNA, NT-v2. Supplementary Data 2 lists the models that significantly outperformed the CNN: None. The CNN's AUC is not printed, so it has no result record; the foundation-model AUCs are in Table 1. (Supplementary Data 1 and Supplementary Data 2, row 'Acceptors')
• reported_finding: amp-feng-20261007-protocol-splice-donor: Reported finding Baseline CNN trained from scratch on Donor: Supplementary Data 1 lists the DNA foundation models the CNN significantly outperformed (one-sided DeLong p<0.01): Caduceus-Ph, DNABERT-2, GROVER, HyenaDNA, NT-v2. Supplementary Data 2 lists the models that significantly outperformed the CNN: None. The CNN's AUC is not printed, so it has no result record; the foundation-model AUCs are in Table 1. (Supplementary Data 1 and Supplementary Data 2, row 'Donors')
• reported_finding: feng-dna-fm-20261013-protocol-t1-coding: Reported finding Baseline CNN trained from scratch on Coding: Supplementary Data 1 lists the DNA foundation models the CNN significantly outperformed (one-sided DeLong p<0.01): DNABERT-2, HyenaDNA, NT-v2. Supplementary Data 2 lists the models that significantly outperformed the CNN: Caduceus-Ph, GROVER. The CNN's AUC is not printed, so it has no result record; the foundation-model AUCs are in Table 1. (Supplementary Data 1 and Supplementary Data 2, row 'Coding')
• reported_finding: feng-dna-fm-20261013-protocol-t1-dnase-i-hypersensitive: Reported finding Baseline CNN trained from scratch on DNase I Hypersensitive: Supplementary Data 1 lists the DNA foundation models the CNN significantly outperformed (one-sided DeLong p<0.01): None. Supplementary Data 2 lists the models that significantly outperformed the CNN: Caduceus-Ph, DNABERT-2. The CNN's AUC is not printed, so it has no result record; the foundation-model AUCs are in Table 1. (Supplementary Data 1 and Supplementary Data 2, row 'DNase I Hypersensitive')
• reported_finding: feng-dna-fm-20261013-protocol-t1-enhancer: Reported finding Baseline CNN trained from scratch on Enhancer: Supplementary Data 1 lists the DNA foundation models the CNN significantly outperformed (one-sided DeLong p<0.01): None. Supplementary Data 2 lists the models that significantly outperformed the CNN: DNABERT-2, NT-v2. The CNN's AUC is not printed, so it has no result record; the foundation-model AUCs are in Table 1. (Supplementary Data 1 and Supplementary Data 2, row 'Enhancer')
• reported_finding: feng-dna-fm-20261013-protocol-t1-enhancer-cohn: Reported finding Baseline CNN trained from scratch on Enhancer Cohn: Supplementary Data 1 lists the DNA foundation models the CNN significantly outperformed (one-sided DeLong p<0.01): HyenaDNA. Supplementary Data 2 lists the models that significantly outperformed the CNN: None. The CNN's AUC is not printed, so it has no result record; the foundation-model AUCs are in Table 1. (Supplementary Data 1 and Supplementary Data 2, row 'Enhancer Cohn')
• reported_finding: feng-dna-fm-20261013-protocol-t1-enhancer-ensembl: Reported finding Baseline CNN trained from scratch on Enhancer Ensembl: Supplementary Data 1 lists the DNA foundation models the CNN significantly outperformed (one-sided DeLong p<0.01): None. Supplementary Data 2 lists the models that significantly outperformed the CNN: None. The CNN's AUC is not printed, so it has no result record; the foundation-model AUCs are in Table 1. (Supplementary Data 1 and Supplementary Data 2, row 'Enhancer Ensembl')
• reported_finding: feng-dna-fm-20261013-protocol-t1-human-tfbs-1: Reported finding Baseline CNN trained from scratch on Human TFBS 1: Supplementary Data 1 lists the DNA foundation models the CNN significantly outperformed (one-sided DeLong p<0.01): None. Supplementary Data 2 lists the models that significantly outperformed the CNN: Caduceus-Ph, GROVER. The CNN's AUC is not printed, so it has no result record; the foundation-model AUCs are in Table 1. (Supplementary Data 1 and Supplementary Data 2, row 'Human TFBS 1')
• reported_finding: feng-dna-fm-20261013-protocol-t1-human-tfbs-2: Reported finding Baseline CNN trained from scratch on Human TFBS 2: Supplementary Data 1 lists the DNA foundation models the CNN significantly outperformed (one-sided DeLong p<0.01): None. Supplementary Data 2 lists the models that significantly outperformed the CNN: Caduceus-Ph, GROVER. The CNN's AUC is not printed, so it has no result record; the foundation-model AUCs are in Table 1. (Supplementary Data 1 and Supplementary Data 2, row 'Human TFBS 2')
• reported_finding: feng-dna-fm-20261013-protocol-t1-human-tfbs-3: Reported finding Baseline CNN trained from scratch on Human TFBS 3: Supplementary Data 1 lists the DNA foundation models the CNN significantly outperformed (one-sided DeLong p<0.01): Caduceus-Ph, DNABERT-2, GROVER, HyenaDNA, NT-v2. Supplementary Data 2 lists the models that significantly outperformed the CNN: None. The CNN's AUC is not printed, so it has no result record; the foundation-model AUCs are in Table 1. (Supplementary Data 1 and Supplementary Data 2, row 'Human TFBS 3')
• reported_finding: feng-dna-fm-20261013-protocol-t1-human-tfbs-4: Reported finding Baseline CNN trained from scratch on Human TFBS 4: Supplementary Data 1 lists the DNA foundation models the CNN significantly outperformed (one-sided DeLong p<0.01): DNABERT-2, HyenaDNA, NT-v2. Supplementary Data 2 lists the models that significantly outperformed the CNN: Caduceus-Ph, GROVER. The CNN's AUC is not printed, so it has no result record; the foundation-model AUCs are in Table 1. (Supplementary Data 1 and Supplementary Data 2, row 'Human TFBS 4')
• reported_finding: feng-dna-fm-20261013-protocol-t1-human-tfbs-5: Reported finding Baseline CNN trained from scratch on Human TFBS 5: Supplementary Data 1 lists the DNA foundation models the CNN significantly outperformed (one-sided DeLong p<0.01): None. Supplementary Data 2 lists the models that significantly outperformed the CNN: Caduceus-Ph, GROVER. The CNN's AUC is not printed, so it has no result record; the foundation-model AUCs are in Table 1. (Supplementary Data 1 and Supplementary Data 2, row 'Human TFBS 5')
• reported_finding: feng-dna-fm-20261013-protocol-t1-open-chromatin-region: Reported finding Baseline CNN trained from scratch on Open chromatin region: Supplementary Data 1 lists the DNA foundation models the CNN significantly outperformed (one-sided DeLong p<0.01): DNABERT-2, GROVER, HyenaDNA, NT-v2. Supplementary Data 2 lists the models that significantly outperformed the CNN: None. The CNN's AUC is not printed, so it has no result record; the foundation-model AUCs are in Table 1. (Supplementary Data 1 and Supplementary Data 2, row 'Open Chromatin Region')
• reported_finding: feng-dna-fm-20261013-protocol-t1-promoter-all-70-bps: Reported finding Baseline CNN trained from scratch on Promoter All 70 bps: Supplementary Data 1 lists the DNA foundation models the CNN significantly outperformed (one-sided DeLong p<0.01): Caduceus-Ph, DNABERT-2, GROVER, HyenaDNA, NT-v2. Supplementary Data 2 lists the models that significantly outperformed the CNN: None. The CNN's AUC is not printed, so it has no result record; the foundation-model AUCs are in Table 1. (Supplementary Data 1 and Supplementary Data 2, row 'Promoter All 70bps')
• reported_finding: feng-dna-fm-20261013-protocol-t1-promoter-all-300-bps: Reported finding Baseline CNN trained from scratch on Promoter All 300 bps: Supplementary Data 1 lists the DNA foundation models the CNN significantly outperformed (one-sided DeLong p<0.01): DNABERT-2, GROVER, HyenaDNA, NT-v2. Supplementary Data 2 lists the models that significantly outperformed the CNN: None. The CNN's AUC is not printed, so it has no result record; the foundation-model AUCs are in Table 1. (Supplementary Data 1 and Supplementary Data 2, row 'Promoter All 300bps')
• reported_finding: feng-dna-fm-20261013-protocol-t1-promoter-gm12878: Reported finding Baseline CNN trained from scratch on Promoter GM12878: Supplementary Data 1 lists the DNA foundation models the CNN significantly outperformed (one-sided DeLong p<0.01): None. Supplementary Data 2 lists the models that significantly outperformed the CNN: Caduceus-Ph, DNABERT-2, GROVER, NT-v2. The CNN's AUC is not printed, so it has no result record; the foundation-model AUCs are in Table 1. (Supplementary Data 1 and Supplementary Data 2, row 'Promoter GM12878')
• reported_finding: feng-dna-fm-20261013-protocol-t1-promoter-hela-s3: Reported finding Baseline CNN trained from scratch on Promoter Hela-S3: Supplementary Data 1 lists the DNA foundation models the CNN significantly outperformed (one-sided DeLong p<0.01): None. Supplementary Data 2 lists the models that significantly outperformed the CNN: Caduceus-Ph, DNABERT-2, GROVER, NT-v2. The CNN's AUC is not printed, so it has no result record; the foundation-model AUCs are in Table 1. (Supplementary Data 1 and Supplementary Data 2, row 'Promoter Hela-S3')
• reported_finding: feng-dna-fm-20261013-protocol-t1-promoter-huvec: Reported finding Baseline CNN trained from scratch on Promoter HUVEC: Supplementary Data 1 lists the DNA foundation models the CNN significantly outperformed (one-sided DeLong p<0.01): None. Supplementary Data 2 lists the models that significantly outperformed the CNN: Caduceus-Ph, DNABERT-2, GROVER, NT-v2. The CNN's AUC is not printed, so it has no result record; the foundation-model AUCs are in Table 1. (Supplementary Data 1 and Supplementary Data 2, row 'Promoter HUVEC')
• reported_finding: feng-dna-fm-20261013-protocol-t1-promoter-nhek: Reported finding Baseline CNN trained from scratch on Promoter NHEK: Supplementary Data 1 lists the DNA foundation models the CNN significantly outperformed (one-sided DeLong p<0.01): HyenaDNA, NT-v2. Supplementary Data 2 lists the models that significantly outperformed the CNN: None. The CNN's AUC is not printed, so it has no result record; the foundation-model AUCs are in Table 1. (Supplementary Data 1 and Supplementary Data 2, row 'Promoter NHEK')
• reported_finding: feng-dna-fm-20261013-protocol-t1-promoter-nontata-70-bps: Reported finding Baseline CNN trained from scratch on Promoter NonTATA 70 bps: Supplementary Data 1 lists the DNA foundation models the CNN significantly outperformed (one-sided DeLong p<0.01): DNABERT-2, GROVER, HyenaDNA, NT-v2. Supplementary Data 2 lists the models that significantly outperformed the CNN: None. The CNN's AUC is not printed, so it has no result record; the foundation-model AUCs are in Table 1. (Supplementary Data 1 and Supplementary Data 2, row 'Promoter NonTATA 70bps')
• reported_finding: feng-dna-fm-20261013-protocol-t1-promoter-nontata-251-bps: Reported finding Baseline CNN trained from scratch on Promoter NonTATA 251 bps: Supplementary Data 1 lists the DNA foundation models the CNN significantly outperformed (one-sided DeLong p<0.01): DNABERT-2, HyenaDNA, NT-v2. Supplementary Data 2 lists the models that significantly outperformed the CNN: None. The CNN's AUC is not printed, so it has no result record; the foundation-model AUCs are in Table 1. (Supplementary Data 1 and Supplementary Data 2, row 'Promoter NonTATA 251bps')
• reported_finding: feng-dna-fm-20261013-protocol-t1-promoter-nontata-300-bps: Reported finding Baseline CNN trained from scratch on Promoter NonTATA 300 bps: Supplementary Data 1 lists the DNA foundation models the CNN significantly outperformed (one-sided DeLong p<0.01): Caduceus-Ph, DNABERT-2, GROVER, HyenaDNA, NT-v2. Supplementary Data 2 lists the models that significantly outperformed the CNN: None. The CNN's AUC is not printed, so it has no result record; the foundation-model AUCs are in Table 1. (Supplementary Data 1 and Supplementary Data 2, row 'Promoter NonTATA 300bps')
• reported_finding: feng-dna-fm-20261013-protocol-t1-promoter-tata-70-bps: Reported finding Baseline CNN trained from scratch on Promoter TATA 70 bps: Supplementary Data 1 lists the DNA foundation models the CNN significantly outperformed (one-sided DeLong p<0.01): Caduceus-Ph, DNABERT-2, GROVER, HyenaDNA, NT-v2. Supplementary Data 2 lists the models that significantly outperformed the CNN: None. The CNN's AUC is not printed, so it has no result record; the foundation-model AUCs are in Table 1. (Supplementary Data 1 and Supplementary Data 2, row 'Promoter TATA 70bps')

Models, benchmarks and other records

Evidence

Source checking verifies the cited claim or transcription. It does not establish independent reproduction.