Unlocking the Regulatory Genome by ARGUS: An Evidence-Constrained Agentic Framework for Interpreting Single Nucleotide Variants

Pratik Dutta; Matthew B. Obusan; Max Chao; Rekha Sathian; Nimisha Papineni; Ramana V. Davuluri

q-bio.GNOct 8, 2026

Unlocking the Regulatory Genome by ARGUS: An Evidence-Constrained Agentic Framework for Interpreting Single Nucleotide Variants

Authors: Pratik Dutta, Matthew B. Obusan, Max Chao, Rekha Sathian, Nimisha Papineni, Ramana V. Davuluri

Organizations: Stony Brook University, Stony Brook, NY, USA · Renaissance School of Medicine at Stony Brook University, Stony Brook, NY, USA

Abstract

Over 90% of disease-associated variants from genome-wide association studies fall in noncoding regulatory regions, yet their functional interpretation remains a central open problem in genomic medicine. Large language models prompted to interpret such variants routinely hallucinate transcription factor (TF) binding changes, fabricate experimental support, and assign biological significance to statistically negligible signals. We present ARGUS (Agentic Regulatory Genomics for an Uncertainty-aware Scientist), which strictly separates deterministic biological computation from LLM-mediated reasoning. ARGUS wraps 458 DNABERT-based TF binding models in a hypothesis-directed investigation loop where a planner selects evidence sources based on current uncertainty, a verifier deterministically interprets each observation, and intermediate results change the investigation path. On variant rs6983267 at the 8q24 cancer risk locus, the same planner produces four divergent trajectories for four TFs. FOXA1 is rescued in 3 steps when real ADASTRA allele-specific binding data (15 experiments, FDR = 0.030) reveals a model false negative masked by saturation. KLF6 traverses 8 steps across ADASTRA, JASPAR motif analysis, and ENCODE cCRE regulatory annotation before abstaining due to mixed indirect evidence. RAD21 abstains in 8 steps after ADASTRA returns a coverage-qualified but nonsignificant allelic test (5 experiments, FDR = 0.65), and SP1, which shares FOXA1's saturated retained prediction, abstains because no direct experimental evidence exists at this locus. All observations come from real ADASTRA, JASPAR, and ENCODE cCRE queries; none are simulated. A comparison of fixed-priority and LLM-mediated planning shows that the LLM planner reaches identical verdicts with fewer tool calls by declining evidence that cannot resolve the claim under test.

Figures & tables

Figure 1: ARGUS architecture. Left: task inputs with precomputed DVR predictions. Center: the observation-dependent investigation loop (state/planner, evidence acquisition, deterministic verifier, stopping rule). Right: claim-specific verification showing how FOXA1 (rescued, 3 steps) and KLF6 (abstained, 8 steps) reach different verdicts from the same planner responding to different intermediate observations. AlphaGenome Atlas remains outside the evaluated loop.

TF	DVR Arm	Steps	Tools Used	Final Status	Key Observation
FOXA1	retained [sat.]	3	ADASTRA	rescued	Significant ASB (FDR = 0.030, 15 exp., 475 reads); contradicts DVR’s non-differential prediction
KLF6	strengthened	8	ADASTRA, JASPAR, cCRE	abstained	ADASTRA unavailable; JASPAR motif opposes DVR ( $\Delta={-}0.147$ ); cCRE supports (pELS); mixed evidence
RAD21	retained [sat.]	8	ADASTRA, JASPAR, cCRE	abstained	Powered but nonsignificant ASB (FDR = 0.65, 5 exp., 104 reads); no RAD21 motif; cCRE context only
SP1	retained [sat.]	8	ADASTRA, JASPAR, cCRE	abstained	No ADASTRA data at locus; motif direction not adjudicated for non-directional arm; cCRE context only

Table 1: ARGUS investigation trajectories for four TFs at rs6983267. Each trajectory is produced by the same planner responding to different intermediate observations.

TF	Variant	Fixed tools	LLM tools	Fixed status	LLM status
FOXA1	rs6983267	A (1/3)	A (1/3)	rescued	rescued
RAD21	rs6983267	A, J, C (3/3)	A, J (2/3)	abstained	abstained
KLF6	rs6983267	A, J, C (3/3)	A, J (2/3)	abstained	abstained
SP1	rs6983267	A, J, C (3/3)	A, J (2/3)	abstained	abstained
BACH1	rs2981578	A, J, C (3/3)	A, J (2/3)	abstained	abstained

Table 2: Fixed vs. LLM planner comparison across five hypotheses on identical frozen evidence snapshots. Tool columns show the sequence each planner selected, with tool calls used out of a budget of 3 in parentheses. LLM results were identical across three independent replicates.

Explore similar work

Jun 5, 2026cs.CL

The Dark Regulome: Disentangling Predictability from Regulation in Genomic Foundation Models

High-grade gliomas integrate into neural circuits through functional synapses with neurons, raising the question of which noncoding elements shape synaptogenic gene expression in tumor cells. The regulatory program written across the dark genome, what we call the

\textit{dark regulome}

, is the natural substrate to probe, and sequence foundation models offer a zero-shot route through in-silico mutagenesis (ISM); yet likelihood-based scoring is tautologically coupled to local sequence predictability, leaving the regulatory interpretation underdetermined. Across three architecturally distinct foundation models (Caduceus-Ph, HyenaDNA, Enformer) and 30,448 dark genome elements at 92 glioma-relevant loci, we introduce a residualization-and-permutation diagnostic that separates predictability-driven from regulation-driven RIS variance. A sharp 10kb proximal-regulatory horizon survives every control we apply, but the LM-derived element-class hierarchy does not: a six-feature linear baseline matches Caduceus top-decile membership at AUC

= 0.985

. Cross-architecture decomposition cleanly separates a sequence-predictability layer (the two language models co-rank long well-predicted transposable elements) from a regulatory-output layer (Enformer alone retains residual cCRE-discriminative signal), with literally zero overlap between the two top-100 lists. Conservation, brain cis-eQTL, and STRING-PPI cross-checks then anchor what biology survives: top-100 elements across all three models are

3.3\times

enriched per model for matching brain eQTLs (

p_\mathrm{emp} < 5\times 10^{-3}

), while a tempting transposable-element regulatory layer and a striking NRXN1+NLGN1 protein-pair convergence both fail proper permutation tests once those tests are constructed. We deliver the diagnostic as a general methodological tool for any ISM-based regulatory study.

Chahat Baranwal, Aaditya Baranwal, Lakshya Nitin Tandon

IIT Jodhpur · University of Central Florida · Northeastern University

Genomics Molecular Representation Learning

Ask PDF

May 29, 2026cs.LG

Position: Genomic Model Research Must Move Beyond Anecdotal Evaluation of Interpretability Methods

Advances in machine learning and computational power have unlocked the predictive potential of the human genome, yet biologists now demand that these models also elucidate the underlying biological mechanisms. While interpretable machine learning (IML) techniques have been increasingly applied to bridge this gap, there has been a pervasive reliance on anecdotal validation: the vast majority of research relies on a single IML method and reports only isolated successful instances. Through a benchmarking study on transcription factor binding, we demonstrate the risks of current practices. We show that different IML methods can often (1) yield contradictory explanations for the same predictions, (2) fail to localize known regulatory motifs, and (3) fail to faithfully reflect the model's internal decision process. In light of this, we argue for a validation framework analogous to clinical trials: just as trials require rigorous design and adverse-event reporting, genomic interpretability must move beyond cherry-picked plausibility toward systematic assessment of consistency, faithfulness, and biological validity. To facilitate this, we propose a tiered framework to guide rigorous evaluation and reporting of genomic IML methods.

Shasha Zhou, Mingyu Huang, Ke Li

Department of Computer Science, University of Exeter, Exeter, UK.

Interpretable ML Genomics Explainability Evaluation

Ask PDF

Jun 6, 2026q-bio.GN

Biological Reasoning-Informed Regression for Interpretable Regulatory DNA Activity Prediction

DNA cis-regulatory elements (CREs) such as enhancers control gene expression levels. Accurately predicting regulatory activity from DNA sequences is valuable but challenging, as it requires understanding complex biological regulatory processes. Existing methods typically regress activity scores from sequences in a black-box manner, limiting both interpretability and regression performance. Meanwhile, large language models (LLMs) benefit from explicit reasoning processes, yet directly applying LLMs to raw DNA sequences performs poorly. In this paper, we bridge this gap by introducing R3LM, a framework that teaches LLMs reasoning-informed regression on regulatory DNA through structured biological knowledge. Specifically, we design a biologically grounded data format that structures DNA's regulatory information for improved LLM understanding, and construct CRE-ReasonBench, the first dataset that associates DNA sequences and activity scores with mechanistic reasoning traces. Through two-stage training that first teaches LLMs reasoning over structured biological information then performs regression, R3LM achieves state-of-the-art performance on enhancer prediction across three cell types, outperforming both LLMs with raw sequence input and specialized DNA models while providing interpretable mechanistic explanations. We expect R3LM as an interpretable reward model that can effectively assist biologists in CRE design. Code is available at https://github.com/DuanYi516/R3LM.

Yi Duan, Zhao Yang, Jiwei Zhu +3

Gaoling School of Artificial Intelligence · Renmin University of China · Beijing, China +1

Interpretable ML

Ask PDF