NVAITC AI Scientist: A Governed End-to-End Research System -- A Hypertension GWAS Case Study

arXiv 2026 66.5 method

TLDR

A governed end-to-end agentic research system for biomedical workflows, validated on a real-world hypertension GWAS with 286,422 individuals.

Reasoning

The paper presents a novel system (NAIS) that integrates planning, orchestration, and human oversight for reproducible research, with strong real-world validation. However, the abstract lacks details on generalizability beyond GWAS and does not address literature review or survey generation.

Read-first score

Read-first score 66.5, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 84.

Recency 8%

100

Uses a gentle age decay so recent papers surface without erasing older foundations. 2026

Topical relevance 42%

Uses existing LLM keyword relevance scores normalized to 0-100. AI scientist,automated scientific discovery,autonomous research agent,automated research,literature review agent,survey generation,automated experimentation,experiment design agent,AI for scientific research,paper writing agent,research automation,scientific discovery agent

Methodology quality 25%

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=analysis,result,validation

Reproducibility 25%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=code,reproduce

Field roles

FrontierMethodology anchor

Rank sensitivity

Stability: volatile; rank range: 48.

Keyword Scores

AI scientist

automated scientific discovery

automated research

automated experimentation

research automation

scientific discovery agent

autonomous research agent

AI for scientific research

experiment design agent

paper writing agent

literature review agent

survey generation

Deep Analysis

Innovations

A governed end-to-end agentic research system (NAIS) that integrates proposal review, execution planning, governed computational routing, reproducible workflow orchestration, evidence generation, and scientist-in-the-loop oversight while keeping data within institutional privacy boundaries.
Demonstration of iterative phenotype refinement through human-AI review in a real-world GWAS, enabling correction of phenotype discrepancies and improvement of the hypertension definition.
Validation of the system by reproducing established hypertension-associated loci and supporting a secondary drug-induced liver injury prediction task, showing outputs comparable to expert-led workflows.

Methodology

NAIS is a domain-general agentic system that orchestrates scientific workflows with governed mechanisms for research planning, data access, and reproducibility. It was validated on a hypertension GWAS using genotype and electronic health record data from 286,422 individuals under an aggregate-only data policy, where the agent planned cohort extraction, executed GWAS, generated quality-control summaries, and drafted publication-oriented outputs. A secondary drug-induced liver injury prediction task used a multimodal graph neural network.

Key Results

The agent-orchestrated GWAS reproduced established hypertension loci (FGF5, ATP2B1, CNNM2, FTO, GRB14) with the strongest signal at FGF5 reaching -log10(p) ~ 70, and the drug-induced liver injury prediction model achieved an AUC of 0.842.

Resources

Paper