DeepVRegulome

462 fine-tuned DNABERT models for regulatory variant effect prediction

DeepVRegulome is an end-to-end framework for predicting the functional impact of small somatic variants in non-coding regulatory regions using fine-tuned DNABERT models. It covers 458 transcription factors and 4 histone modifications from ENCODE ChIP-seq data, validated against Yan et al. (2021) SNP-SELEX experimental variant-effect measurements.

Resource	Link
Paper	arXiv:2511.09026
Code	GitHub: DavuluriLab/DeepVRegulome
PyPI	pip install deepvregulome
Web App	deepvregulome.streamlit.app

Key Features

• 462 fine-tuned models (458 TF-binding + 4 histone mark) trained on ENCODE ChIP-seq peaks
• Variant effect scoring via log-odds ratio between reference and alternate alleles
• Attention-based motif analysis for interpretable predictions
• Experimentally validated against SNP-SELEX (Yan et al., 2021, Nature Genetics): mean per-TF AUROC = 0.611 across 61 evaluable TFs
• Benchmarked against DeepSEA, Enformer, Borzoi, and AlphaGenome

Installation

pip install deepvregulome

Quick Start (with deepvregulome package)

import deepvregulome as dvr

# Initialize
pipeline = dvr.DVR()

# List all 462 available models
models = pipeline.list_models()
print(f"{len(models)} models available")

# Score a variant
results = pipeline.score_variant(
    chrom="chr1", pos=1000000, ref="A", alt="G",
    models=["CTCF", "SP1", "MYC"]
)
print(results)

Quick Start (direct transformers usage)

from transformers import AutoModelForSequenceClassification, AutoTokenizer
import torch

# Load any model using subfolder
model_name = "CTCFL"  # or "SP1", "MYC", "H3K27ac", etc.
tokenizer = AutoTokenizer.from_pretrained(
    "duttaprat/DeepVRegulome", subfolder=f"models/{model_name}"
)
model = AutoModelForSequenceClassification.from_pretrained(
    "duttaprat/DeepVRegulome", subfolder=f"models/{model_name}"
)
model.eval()

# Convert DNA to 6-mer representation
def to_kmer(seq, k=6):
    return " ".join([seq[i:i+k] for i in range(len(seq) - k + 1)])

# Predict binding probability
sequence = "ATCGATCG..."  # 301bp DNA sequence
inputs = tokenizer(to_kmer(sequence), return_tensors="pt",
                   max_length=512, truncation=True, padding=True)
with torch.no_grad():
    prob = torch.softmax(model(**inputs).logits, dim=-1)[0][1].item()
print(f"{model_name} binding probability: {prob:.4f}")

Variant Effect Scoring

import math

def score_variant(model, tokenizer, ref_seq, alt_seq):
    probs = {}
    for name, seq in [("REF", ref_seq), ("ALT", alt_seq)]:
        inputs = tokenizer(to_kmer(seq), return_tensors="pt",
                           max_length=512, truncation=True, padding=True)
        with torch.no_grad():
            probs[name] = torch.softmax(model(**inputs).logits, dim=-1)[0][1].item()

    eps = 1e-7
    lo_ref = math.log((probs["REF"] + eps) / (1 - probs["REF"] + eps))
    lo_alt = math.log((probs["ALT"] + eps) / (1 - probs["ALT"] + eps))
    return {
        "prob_ref": probs["REF"],
        "prob_alt": probs["ALT"],
        "log_odds_change": lo_alt - lo_ref,
        "disrupted": abs(lo_alt - lo_ref) > 2.0,
    }

Available Models (462)

Each model is stored in a models/<NAME>/ subfolder. Load with:

AutoModel.from_pretrained("duttaprat/DeepVRegulome", subfolder="models/<NAME>")

Transcription Factor Models (458)

Model	Accuracy	F1	ROC-AUC	PR-AUC	ChIP-seq Peaks
CTCFL	98.39	98.4	99.71	99.7	12,743
ZNF426	97.09	96.64	98.08	98.65	7,915
SAFB	97.02	96.81	98.38	98.57	5,361
RBM34	97.0	96.53	98.18	98.44	928
TAF15	96.74	96.41	99.02	98.58	17,900
PCBP1	96.7	95.94	99.19	98.66	16,229
SRSF3	96.6	96.52	98.2	95.53	1,610
RBM14	96.57	95.73	98.8	98.4	2,463
KDM4A	96.07	95.99	99.08	98.72	25,968
NFYA	95.91	95.19	96.88	93.73	4,313
SPI1	95.79	95.78	98.75	98.38	104,150
HLF	95.71	95.83	98.88	98.74	66,792
EGR2	95.68	99.22	99.75	99.6	57,610
HNRNPLL	95.43	93.73	98.77	97.88	25,270
HNRNPK	95.41	94.27	97.05	96.06	24,014
FIP1L1	95.35	94.16	98.25	98.13	10,123
ZNF654	95.31	95.28	97.93	98.17	33,357
ZNF146	95.23	94.66	97.57	97.69	38,167
E2F6	95.14	93.71	98.25	96.91	30,723
C11orf30	95.13	94.98	98.63	98.61	58,686
FUS	95.03	93.76	94.87	93.76	7,495
USF1	94.95	94.93	98.05	97.78	60,863
ZNF770	94.91	95.01	98.04	97.28	57,572
AGO1	94.81	92.98	97.19	96.16	16,201
MAFG	94.74	94.7	97.91	97.78	44,030
ZFHX2	94.73	99.21	99.82	99.74	61,052
THAP1	94.7	94.48	98.22	97.82	5,210
HES2	94.55	84.34	97.15	84.39	6,520
SRSF1	94.49	93.59	98.34	97.9	9,933
NFE2	94.22	99.19	99.76	99.72	53,717
MITF	94.14	93.57	98.35	98.12	36,171
MAFF	93.91	93.91	97.86	97.17	65,684
PCBP2	93.89	93.23	96.53	96.61	5,328
ZIC2	93.77	93.9	97.82	96.94	56,766
GLIS2	93.74	93.14	98.22	97.8	33,566
ZNF121	93.64	98.86	99.34	99.26	35,805
SP2	93.6	93.13	96.81	95.36	27,956
SCRT2	93.59	98.97	99.56	99.17	59,462
BATF	93.58	93.82	98.16	97.93	36,658
HNRNPL	93.57	92.29	97.59	96.88	24,447
NFE2L2	93.53	91.27	97.24	96.09	26,392
ZNF585B	93.43	93.4	97.78	98.0	7,381
NFYB	93.31	91.7	96.3	95.22	14,696
ZSCAN4	93.25	92.72	97.64	97.75	23,997
ZNF316	93.22	93.43	97.42	97.55	97,518
ZNF140	93.19	92.79	96.45	96.93	7,623
ZNF777	93.12	92.58	97.24	97.07	7,181
CBFA2T2	93.06	93.09	97.31	97.25	32,239
ATF4	93.0	92.84	97.72	97.64	39,441
TFAP4	93.0	92.84	96.79	94.97	27,035
CTBP2	92.98	92.36	96.78	94.11	7,351
E2F1	92.88	86.4	95.25	91.61	18,927
MXD3	92.88	92.08	98.24	97.87	14,677
ZNF202	92.87	92.87	96.29	96.5	6,476
ZNF266	92.84	92.56	96.93	96.93	5,097
ZNF263	92.83	91.44	97.39	96.16	35,011
ZNF433	92.77	92.49	96.97	97.7	3,530
RBM39	92.73	91.61	97.44	96.77	29,478
ZNF680	92.69	92.24	97.36	97.68	10,483
ZNF341	92.66	92.24	97.24	96.26	33,413
E2F4	92.64	90.76	97.32	97.07	11,270
CBFA2T3	92.59	92.67	96.97	96.03	51,994
ZNF555	92.57	92.23	94.78	96.56	6,664
RBM25	92.56	92.48	96.96	97.05	47,193
MAFK	92.55	92.65	97.36	97.6	121,216
ZNF623	92.53	92.19	96.77	97.51	14,353
KAT2A	92.45	91.3	92.39	92.84	171
SIN3B	92.45	90.29	97.21	96.89	11,700
CEBPA	92.39	98.89	99.7	99.6	57,842
U2AF2	92.37	91.72	96.2	94.8	2,777
ZBTB10	92.27	91.31	96.02	94.41	15,441
PBX3	92.25	92.23	96.51	95.35	12,981
SAP30	92.08	90.13	96.21	92.3	15,727
PHF8	92.03	88.66	96.25	95.67	30,275
PAX5	92.02	90.77	96.81	94.42	36,090
ZNF785	92.01	91.35	97.35	97.16	6,605
ZNF76	92.0	90.95	93.83	89.14	11,676
HNF4G	91.93	90.87	96.43	95.82	35,204
AHR	91.82	91.39	97.21	96.43	10,611
ZC3H11A	91.8	91.33	96.31	97.12	9,905
RFX1	91.79	90.4	96.44	96.43	38,873
SMAD4	91.75	91.7	96.19	94.92	43,342
UBTF	91.73	89.06	94.72	91.45	19,173
ZNF704	91.71	90.84	94.66	94.52	1,834
CEBPB	91.69	98.47	99.54	99.42	148,540
TAL1	91.69	90.15	96.07	95.03	36,574
ZBTB7A	91.69	90.22	97.14	95.99	37,497
NFIL3	91.67	91.52	96.42	96.0	38,852
XRCC5	91.61	90.04	95.9	93.22	32,847
U2AF1	91.6	89.92	96.08	96.0	8,379
ZNF34	91.58	91.34	95.28	96.71	9,650
FOXP2	91.57	90.78	96.14	93.84	24,150
AGO2	91.55	89.26	95.16	94.66	33,079
RUNX3	91.53	98.64	99.5	99.27	69,903
ZBED5	91.49	91.43	97.14	96.94	5,100
RARA	91.46	91.59	95.97	94.89	42,489
FOS	91.45	98.69	99.62	99.63	169,317
SP3	91.44	90.22	95.27	94.77	23,619
NANOG	91.43	90.45	95.72	94.36	15,681
USF2	91.39	89.66	96.88	96.27	37,240
HNF4A	91.37	91.63	96.12	95.11	89,892
SP5	91.37	91.12	95.54	94.28	23,155
ZNF354C	91.37	90.59	95.08	96.39	1,640
ZNF444	91.35	90.8	95.64	94.38	25,366
GMEB2	91.33	90.21	95.76	95.89	4,800
ZNF677	91.33	91.35	94.68	94.52	5,777
ETV5	91.18	90.89	96.55	95.11	29,604
IRF4	91.18	91.14	96.12	95.26	21,810
PRPF4	91.14	89.78	96.21	95.93	8,605
FOXA3	91.12	91.46	96.8	96.6	45,297
ZNF292	91.11	90.59	94.63	96.11	1,900
GLIS1	91.07	90.1	96.04	95.32	58,506
FOXA2	91.05	91.28	96.39	96.06	99,343
KLF9	90.99	87.95	95.59	94.68	32,290
SOX5	90.98	90.85	95.23	93.77	37,979
BCL6	90.97	90.9	96.3	95.48	38,142
IRF3	90.97	87.65	96.01	95.01	5,726
ZNF223	90.97	90.6	95.52	95.63	5,176
HCFC1	90.96	85.32	94.4	91.32	19,700
DEAF1	90.85	90.18	94.68	95.53	3,170
HNRNPH1	90.85	89.75	96.02	95.65	2,446
CEBPZ	90.82	86.79	89.86	85.64	1,971
PTBP1	90.8	89.88	94.48	89.35	8,063
STAT3	90.79	90.81	96.82	96.85	65,231
KAT2B	90.77	90.48	96.6	96.8	3,106
SOX13	90.76	90.83	96.32	95.69	48,796
ZNF423	90.75	90.32	94.63	93.05	11,299
STAT2	90.73	89.46	95.84	93.72	4,303
MEIS2	90.71	90.78	96.27	96.07	52,199
ATF2	90.69	90.57	94.77	94.25	110,897
NFIA	90.67	90.58	95.6	95.07	27,816
PATZ1	90.67	89.53	96.62	95.93	35,416
ZNF398	90.66	89.94	94.96	92.96	26,224
KLF1	90.65	89.32	94.52	92.79	40,024
NFIB	90.65	87.93	96.05	95.1	27,982
SCRT1	90.65	90.42	94.93	94.54	25,487
PLRG1	90.64	90.06	96.68	96.85	2,256
PRDM1	90.62	95.17	98.83	98.73	50,489
ZNF133	90.61	90.22	95.36	94.71	8,423
EBF1	90.59	90.64	95.31	94.3	56,947
SREBF2	90.59	89.11	96.2	95.55	4,933
THAP11	90.57	89.79	94.85	93.4	31,698
POU5F1	90.56	89.71	96.35	95.35	7,870
ZNF48	90.56	89.58	94.31	91.0	28,840
RB1	90.5	87.95	94.0	91.78	31,314
HNF1A	90.44	90.13	93.47	91.12	12,818
ZBTB6	90.44	90.06	95.43	95.01	13,955
KDM5B	90.43	88.75	96.53	95.91	18,356
GLI4	90.37	90.04	96.45	96.57	6,130
RERE	90.37	89.28	95.15	93.11	11,849
ZNF521	90.37	90.04	96.17	96.82	1,585
FOXP1	90.33	89.44	95.14	93.99	31,380
ZBTB26	90.33	87.38	93.76	90.92	32,837
ZNF596	90.31	89.88	95.56	95.65	14,087
ZBTB21	90.29	89.52	96.14	95.86	18,547
HMGXB4	90.27	89.5	93.94	91.67	25,984
TEAD1	90.27	90.29	95.53	94.84	33,326
SFPQ	90.18	90.09	91.7	90.66	335
ZNF660	90.18	89.77	94.75	94.22	29,821
ZNF658	90.14	88.6	93.92	94.54	1,325
SKI	90.12	89.27	95.47	93.65	30,460
SRF	90.12	87.92	96.0	94.95	27,692
CEBPG	90.11	90.14	96.25	96.42	87,912
ZHX1	90.1	88.61	94.55	92.57	5,512
RUNX1	90.07	89.3	94.29	91.85	7,554
RXRB	90.07	89.83	94.62	92.92	38,236
ZBTB48	90.05	89.29	96.26	95.22	24,899
ZNF747	90.02	88.84	95.87	95.94	1,427
VEZF1	90.01	88.14	94.32	93.48	35,726
GATA1	89.96	87.31	93.39	91.08	33,326
ZNF670	89.95	88.3	95.12	95.63	1,288
ARID4B	89.94	88.99	94.05	92.27	42,648
ZNF449	89.94	89.32	94.61	94.5	18,374
KLF17	89.93	89.05	96.18	95.48	25,375
ZBTB20	89.91	88.23	95.2	92.94	31,889
ELF3	89.85	89.43	95.06	93.86	35,966
ZMIZ1	89.85	90.09	95.22	95.28	1,077
ZNF16	89.85	89.14	95.41	95.27	2,784
ZNF143	89.84	98.49	99.48	99.5	66,295
ZSCAN5A	89.84	89.2	93.59	90.92	8,202
ZNF513	89.78	88.85	94.49	95.41	11,211
PPARG	89.77	89.47	94.59	93.53	23,903
ZBTB8A	89.73	88.24	94.0	92.72	32,420
IRF5	89.71	89.41	93.56	94.62	2,755
ZMYM3	89.71	89.32	95.52	95.45	43,157
DEK	89.7	88.82	95.38	93.29	8,490
GABPB1	89.7	89.12	94.29	92.92	47,437
ZNF629	89.69	88.91	94.14	92.43	37,813
ETV1	89.68	89.39	95.3	95.09	20,672
ZNF394	89.67	88.63	94.54	94.01	28,297
TEAD4	89.66	89.49	95.4	94.56	34,189
PHF20	89.63	88.59	94.4	91.57	13,930
NFYC	89.59	89.07	95.3	95.46	15,080
RBBP5	89.59	85.4	96.01	94.46	30,322
ELK1	89.58	84.66	91.33	89.55	12,298
HMBOX1	89.55	88.79	95.65	94.82	24,883
KDM5A	89.53	82.52	92.86	90.18	13,023
MEF2B	89.51	88.5	94.65	93.45	34,822
WHSC1	89.51	89.25	94.42	89.89	1,801
ZNF692	89.51	88.62	94.15	92.63	30,183
SAP130	89.46	89.63	93.52	92.26	58,021
TCF7L2	89.46	86.81	94.06	91.99	28,567
SETDB1	89.45	87.69	96.29	95.41	4,903
ZKSCAN1	89.45	87.61	94.37	92.44	20,226
EGR1	89.43	89.26	93.84	93.58	67,451
GATA4	89.43	88.99	94.52	91.79	12,156
SUZ12	89.36	82.88	94.21	91.33	24,316
HDAC6	89.35	87.38	95.07	94.75	5,762
TEAD3	89.35	89.2	95.32	93.94	42,854
SRSF9	89.34	87.07	96.1	95.7	846
MLX	89.33	88.85	93.58	91.33	14,640
ARID1B	89.32	88.77	95.17	94.59	45,992
ZSCAN30	89.31	88.65	94.12	92.49	24,704
ZFP36	89.3	86.44	94.32	92.7	32,993
ZFP64	89.3	88.8	94.25	93.12	9,433
ZNF518A	89.3	88.1	94.68	95.04	16,690
ZNF512	89.29	87.52	95.55	95.6	17,500
MEF2C	89.25	89.19	94.66	93.88	11,421
ARID2	89.24	88.78	93.47	91.82	11,585
IKZF5	89.21	88.59	94.7	92.99	25,386
PRDM2	89.19	88.58	93.29	91.23	3,824
KDM6A	89.17	88.82	95.31	94.56	12,193
RXRA	89.17	89.44	94.78	93.82	85,611
MEF2A	89.15	89.07	95.08	95.07	22,725
NRF1	89.15	87.63	93.52	92.27	41,391
SREBF1	89.15	87.65	95.62	95.11	13,518
BCOR	89.08	88.74	95.34	94.44	41,985
FOSL1	89.06	88.05	94.11	93.96	43,596
ZFP69B	89.05	88.39	94.54	93.79	21,782
ZNF837	89.0	88.53	94.06	94.52	2,332
CREB1	88.96	89.04	95.48	96.05	89,914
KDM3A	88.94	88.07	92.68	88.4	15,970
POU2F2	88.92	88.77	93.25	89.71	19,963
RBPJ	88.92	87.6	94.85	93.61	24,864
EED	88.9	88.18	94.22	92.87	33,175
IRF2	88.9	87.49	93.77	92.14	31,753
CBFB	88.89	87.72	93.81	91.26	18,476
ZNF362	88.88	87.44	94.25	92.46	20,426
HMG20A	88.87	88.31	94.68	93.45	24,288
NR2F6	88.87	88.87	93.72	92.0	59,262
MIXL1	88.8	88.4	94.17	93.38	25,659
ZNF768	88.79	88.37	92.77	94.22	7,605
ZNF791	88.77	87.99	94.79	95.45	3,900
GTF2B	88.76	86.89	94.01	94.5	1,890
ZNF652	88.76	88.46	95.01	94.59	15,412
CBX1	88.75	87.21	94.69	92.62	12,838
KLF4	88.72	86.04	92.03	91.82	6,883
HHEX	88.7	88.36	95.48	94.32	7,808
ZFP91	88.7	87.92	93.44	93.1	12,862
ETS1	88.69	84.7	94.07	90.02	30,554
FOSL2	88.66	88.52	92.87	91.8	54,934
GMEB1	88.61	86.89	94.78	94.21	22,958
MYRF	88.61	88.08	92.66	88.81	6,038
CREB3L1	88.6	86.11	93.55	91.45	16,274
MBD2	88.6	85.96	94.29	92.38	20,596
SP7	88.58	87.43	93.73	92.33	43,425
ZNF384	88.58	88.42	94.86	94.74	49,451
ZNF664	88.58	88.58	94.23	93.8	26,774
AEBP2	88.57	88.27	92.65	93.95	2,439
HOMEZ	88.57	88.43	94.49	93.05	25,047
DRAP1	88.55	87.7	93.04	91.43	23,241
ZNF514	88.55	88.75	93.62	94.15	1,637
HSF1	88.53	86.77	93.77	93.41	3,588
FOXA1	88.52	88.77	94.82	94.45	144,475
KLF10	88.52	87.13	93.65	91.19	18,933
ZNF18	88.49	87.85	93.25	92.44	18,879
GATAD1	88.45	87.54	92.76	90.62	22,354
BRCA1	88.43	84.43	91.88	90.63	4,271
GATA3	88.41	88.76	93.95	93.68	89,323
TBX3	88.37	86.69	92.33	87.32	16,097
HIC1	88.34	87.87	95.07	94.7	24,393
ELF4	88.29	85.91	91.81	90.4	19,552
KLF7	88.28	87.47	95.07	95.1	11,741
HBP1	88.25	87.02	94.35	92.31	13,101
MAZ	88.25	88.01	93.3	92.34	57,653
RLF	88.25	87.22	91.97	89.17	9,482
YY2	88.25	87.53	94.5	93.99	9,753
ZNF561	88.25	87.47	92.52	89.34	21,469
PBX2	88.24	86.87	92.57	90.17	38,883
ESRRA	88.17	86.95	93.13	89.98	41,619
GATA2	88.17	88.14	93.72	92.79	71,662
ZNF189	88.17	87.09	93.99	92.74	35,323
MIER2	88.16	87.72	93.74	92.28	15,598
DMAP1	88.15	86.04	92.88	89.99	21,794
TBX21	88.15	87.4	94.73	94.2	38,025
RFX3	88.14	87.7	92.23	93.25	6,180
FEZF1	88.12	87.4	94.8	93.99	31,350
ZNF248	88.12	87.31	92.8	92.85	2,499
ZNF600	88.11	87.42	93.97	90.88	45,001
NR2F2	88.09	88.0	95.03	94.56	55,514
ZNF560	88.07	87.67	95.15	96.12	3,447
BHLHE40	88.01	88.0	92.37	91.72	90,623
NFKBIZ	88.0	86.89	93.2	90.86	12,731
SOX6	87.97	87.18	92.97	91.69	36,530
NCOA1	87.95	83.89	92.56	89.24	26,193
ZNF2	87.94	86.44	92.68	91.0	24,945
JUNB	87.93	88.02	92.72	91.19	51,972
KDM4B	87.93	86.42	93.56	94.18	7,803
ZNF10	87.88	87.16	93.83	93.93	17,922
MYB	87.86	86.65	94.72	93.1	3,337
MGA	87.84	86.5	91.45	85.5	28,338
ZBTB2	87.82	86.59	94.58	94.15	16,341
ZEB1	87.82	85.07	94.37	93.15	20,733
REPIN1	87.81	86.18	94.22	94.58	5,721
ZNF148	87.79	85.69	92.13	92.12	21,902
ERF	87.78	86.48	92.92	90.58	27,611
KLF8	87.78	86.2	91.63	88.66	27,450
MYC	87.77	87.53	93.73	92.94	94,483
TRIM22	87.77	97.99	99.03	99.13	55,932
RFX5	87.76	81.96	92.33	89.55	24,728
DACH1	87.72	87.31	92.88	92.63	18,904
ZNF239	87.71	86.87	92.62	91.05	4,689
ZNF366	87.69	97.9	99.1	98.73	44,505
PRDM6	87.65	87.25	94.46	92.98	42,954
ATF3	87.61	87.36	92.96	92.73	94,190
RAD51	87.58	84.45	93.59	92.92	36,133
PRDM4	87.55	87.06	93.45	92.04	20,853
KAT8	87.54	85.89	92.39	89.64	31,059
MBD1	87.54	84.57	92.74	88.6	18,264
PKNOX1	87.54	87.54	94.59	94.36	113,107
WT1	87.54	86.69	92.85	89.27	26,226
CHD2	87.52	86.08	93.75	91.52	41,627
ZXDB	87.51	86.1	91.92	88.84	30,796
HMG20B	87.5	86.97	93.22	90.82	12,656
SIX5	87.49	79.23	89.33	86.03	10,595
MAX	87.44	87.59	92.05	89.23	94,496
TRIM24	87.42	84.62	93.51	90.72	35,639
NONO	87.4	82.54	89.28	84.68	29,192
NFE2L1	87.36	83.51	90.34	91.43	10,261
ZNF391	87.36	86.05	92.69	90.97	12,610
ZSCAN16	87.36	87.06	92.03	89.65	6,915
GATAD2A	87.27	87.34	92.84	92.14	59,253
SMARCA4	87.23	87.23	92.59	91.47	94,655
SMC3	87.23	86.92	93.67	94.33	89,498
ZNF282	87.23	85.73	92.1	90.4	12,994
ZNF645	87.19	85.99	93.82	93.71	3,597
SIRT6	87.18	86.27	91.81	89.63	831
ZBTB11	87.17	83.11	93.24	92.37	34,307
ZNF205	87.17	85.68	93.89	92.45	17,444
KLF13	87.15	86.93	90.16	85.26	11,590
NKRF	87.15	82.95	93.67	92.25	27,626
OSR2	87.15	86.49	92.7	90.29	33,243
CC2D1A	87.14	86.24	92.24	91.48	24,056
NFIC	87.14	87.09	92.88	93.61	74,803
ZNF138	87.12	86.79	93.04	94.92	1,611
PHB2	87.11	86.36	93.7	93.49	8,362
ZHX2	87.11	84.59	92.57	90.95	14,751
ZBTB44	87.09	86.97	91.7	89.35	19,961
ZNF614	87.09	86.34	92.96	90.43	21,841
ZNF501	87.07	84.58	89.56	85.25	18,650
ZNF547	87.07	87.18	93.08	93.55	4,777
E4F1	87.05	83.51	91.86	88.4	27,783
ZNF530	87.0	86.21	92.54	93.54	2,193
MYNN	86.97	85.32	92.29	90.37	19,348
INSM2	86.94	86.19	92.8	91.74	16,385
ZBTB7B	86.94	85.85	93.93	93.22	9,156
BCL11B	86.93	86.01	92.01	89.38	11,525
ZBTB12	86.93	86.23	92.9	91.71	10,363
PML	86.92	85.32	93.87	93.51	18,068
SALL2	86.9	86.55	93.23	94.22	2,489
NR2F1	86.87	86.87	93.08	92.05	70,026
LCORL	86.86	86.86	93.26	93.83	10,617
THRB	86.86	86.05	92.7	91.63	14,895
ZGPAT	86.84	85.13	90.76	86.79	33,522
MIER3	86.83	85.69	92.53	91.2	19,565
KMT2B	86.81	84.82	92.77	91.54	17,871
TGIF2	86.8	85.45	92.88	90.99	18,162
IKZF2	86.79	86.64	92.67	91.71	52,015
ZBTB33	86.78	92.02	97.16	96.92	96,111
NEUROD1	86.73	85.61	92.99	91.71	19,165
ATF7	86.72	86.65	92.77	93.41	105,695
ELF1	86.71	86.32	93.8	94.5	70,684
RELB	86.71	84.55	93.6	91.95	36,669
SMARCE1	86.71	86.36	92.37	91.4	48,505
ZNF843	86.68	85.41	92.05	89.55	22,911
ATM	86.67	85.99	93.74	94.31	2,162
ZNF473	86.67	86.34	92.54	93.0	2,252
MXD4	86.63	84.43	91.34	88.33	18,180
CREM	86.61	86.56	92.96	93.26	72,466
ZNF707	86.61	86.66	93.38	93.31	3,329
ZNF157	86.6	86.52	89.54	86.21	3,778
IKZF3	86.59	85.8	91.99	88.04	29,535
DIDO1	86.58	85.66	92.29	92.48	7,186
ZSCAN26	86.56	85.08	93.43	94.67	2,537
MXI1	86.53	86.12	94.05	94.26	52,982
ZSCAN18	86.51	85.15	92.3	92.92	4,130
CDC5L	86.47	85.08	93.59	93.67	5,151
KLF5	86.44	85.35	92.7	90.15	11,680
ZNF512B	86.43	83.44	93.26	91.76	8,552
LARP7	86.42	84.12	93.32	92.81	27,160
ZNF280C	86.42	86.07	92.27	92.66	1,292
L3MBTL2	86.41	86.23	91.05	89.82	52,245
MCM2	86.41	85.41	93.45	94.29	4,549
E2F7	86.4	85.34	91.15	91.4	2,985
ZNF404	86.4	85.45	91.88	93.7	2,249
MTA1	86.34	84.24	93.57	92.13	22,370
GTF2F1	86.32	83.54	92.14	90.16	39,350
ZNF324	86.32	84.57	93.81	93.44	14,713
RAD21	86.3	97.92	99.11	98.99	188,052
NR2C1	86.27	81.88	90.51	85.79	31,898
RBAK	86.25	85.74	92.56	93.16	3,642
TCF7	86.23	85.39	92.08	90.33	30,462
ZNF776	86.23	85.19	92.22	92.04	2,838
ARHGAP35	86.18	85.43	93.57	93.39	4,817
STAT1	86.15	83.28	92.57	90.18	11,457
ZNF621	86.14	85.57	92.51	92.93	1,976
YBX3	86.12	84.54	90.9	90.72	2,599
ZNF610	86.07	84.35	91.36	89.32	13,333
GLI2	86.05	86.28	91.17	92.26	5,650
KLF6	86.05	84.14	91.14	87.94	15,409
OVOL3	86.05	85.14	90.46	88.66	9,089
ASH1L	85.96	84.17	92.97	93.93	7,269
KLF16	85.91	85.24	90.83	90.08	48,641
NR3C1	85.91	85.89	93.29	93.1	62,779
ZNF114	85.91	82.83	91.36	91.57	1,444
PRDM10	85.9	86.39	91.62	88.42	68,962
ZSCAN29	85.87	82.16	90.3	85.84	21,597
IRF1	85.84	80.68	89.75	87.82	32,357
ZSCAN9	85.83	84.87	91.14	89.95	17,526
ZFP3	85.82	84.76	93.06	93.73	4,550
MNT	85.79	85.08	92.76	93.16	83,589
TAF7	85.79	81.76	93.28	92.37	21,002
MZF1	85.78	85.16	91.67	89.93	17,643
ZFP37	85.78	83.76	89.27	86.32	16,553
PHF21A	85.74	85.18	88.24	82.19	8,878
TOE1	85.74	82.45	89.5	89.69	29,556
BACH1	85.62	83.64	90.78	88.15	41,151
ZNF7	85.61	84.92	92.67	92.68	8,038
ZNF211	85.59	84.6	91.95	92.44	1,919
ZSCAN23	85.58	84.31	92.3	91.5	8,505
FOXM1	85.57	83.72	90.82	89.44	20,757
SUPT5H	85.56	81.99	93.26	92.1	22,469
BCL3	85.55	83.8	93.27	92.67	43,282
TFDP1	85.52	80.8	89.14	84.38	30,292
TAF9B	85.46	83.58	87.02	79.88	13,069
ZNF792	85.46	83.82	92.95	91.81	24,699
JUND	85.43	85.33	93.1	93.42	166,033
SIN3A	85.43	85.07	93.29	93.11	65,597
BCL11A	85.41	84.36	92.15	91.13	33,663
NR2C2	85.39	80.24	89.73	86.9	29,811
POLR2G	85.39	84.83	92.98	93.64	59,528
ZNF311	85.34	84.08	92.44	93.57	2,206
ZNF571	85.34	84.53	92.7	93.55	1,488
EP400	85.29	82.21	89.96	88.24	25,324
ZFP1	85.26	83.98	91.92	90.8	12,129
RCOR2	85.25	84.38	90.5	88.89	19,532
REST	85.25	84.97	90.84	91.49	119,743
MCM3	85.24	84.32	93.0	93.26	2,561
SP1	85.24	84.86	91.19	91.14	120,925
PHF5A	85.23	82.4	90.52	88.55	9,570
ZNF169	85.23	84.36	91.23	90.6	2,873
TSHZ1	85.17	84.45	92.39	91.76	11,804
MTA2	85.15	84.91	90.5	89.57	59,001
GFI1B	85.11	84.57	91.26	90.52	10,694
ZNF558	85.1	84.16	91.1	91.2	13,849
ZNF19	85.06	84.17	91.34	93.34	1,998
SMAD5	85.03	79.12	92.23	90.37	22,402
SNRNP70	85.02	82.87	89.28	89.69	871

Histone Modification Models (4)

Model	Accuracy	F1	ROC-AUC	PR-AUC	ChIP-seq Peaks
H4K12ac	89.57	89.53	96.1	95.46	42,837
H2AK9ac	88.91	89.24	94.58	92.7	129,243
H3K9me1	85.43	87.45	93.75	91.95	196,657
H3K23me2	85.39	85.45	93.76	93.34	77,099

Model Architecture

All models are fine-tuned from the DNABERT pretrained checkpoint (6-new-12w-0), which uses a 6-mer tokenization of DNA sequences. Each model is a BertForSequenceClassification with 2 output classes (bound / unbound).

• Base model: DNABERT (Ji et al., 2021, Bioinformatics)
• Input: 6-mer tokenized DNA sequences (301bp)
• Training data: ENCODE ChIP-seq peaks (positive) vs. length/GC-matched genomic background (negative)
• Negative sequences: Generated via MMseqs2 clustering at 80% identity threshold to ensure sequence diversity

Experimental Validation

DeepVRegulome was validated against the Yan et al. (2021) SNP-SELEX dataset (GSE118725, Nature Genetics), which provides experimentally measured allelic effects of genetic variants on transcription factor binding for 270 TFs.

• 61 TFs overlap between DeepVRegulome's 458 TF models and SNP-SELEX
• Mean per-TF AUROC: 0.611
• 53/61 TFs achieve BH-corrected significance at q < 0.05
• DeepVRegulome significantly outperforms DeepSEA (Wilcoxon p = 3.1 x 10^-5)
• Performance is statistically indistinguishable from Enformer (p = 0.353) and Borzoi (p = 0.329)

Citation

If you use DeepVRegulome, please cite:

@article{dutta2025deepvregulome,
  title={DeepVRegulome: Deep Learning Predicts Functional Impact of Short Genomic 
         Variants on the Human Regulome with Application to Cancer},
  author={Dutta, Pratik and Obusan, Matthew and Sathian, Rekha and Davuluri, Ramana V.},
  journal={arXiv preprint arXiv:2511.09026},
  year={2025}
}

License

CC-BY-NC-4.0