From Literature to Hypotheses: An AI Co-Scientist System for Biomarker-Guided Drug Combination Hypothesis Generation

Raneen Younis; Suvinava Basak; Lukas Chavez; Zahra Ahmadi

cs.CLFeb 28, 2026

From Literature to Hypotheses: An AI Co-Scientist System for Biomarker-Guided Drug Combination Hypothesis Generation

Authors: Raneen Younis, Suvinava Basak, Lukas Chavez, Zahra Ahmadi

Organizations: Peter L. Reichertz Institute for Medical Informatics (PLRI), Hannover Medical School · Lower Saxony Center for Artificial Intelligence and Causal Methods in Medicine (CAIMed) · Sanford Burnham Prebys (SBP) Medical Discovery Institute

Abstract

The rapid growth of biomedical evidence makes it difficult to translate biomarker mechanisms into actionable drug combination hypotheses. We present CoDHy, an interactive AI co-scientist for biomarker-guided hypothesis generation in oncology. CoDHy constructs task-specific knowledge graphs from curated databases and biomedical literature, then combines graph embeddings with agent-based reasoning to generate, validate, and rank evidence-grounded drug combinations. Through a web interface, researchers specify the biomarker, cancer context, and literature scope; inspect supporting evidence and intermediate results; and iteratively refine the generated hypotheses. The demonstration presents CoDHy's end-to-end workflow and shows how researchers can interactively explore and compare mechanistically supported drug combinations while remaining in control of hypothesis prioritization.

Figures & tables

Fig. 1 : CoDHy workflow. Users specify a biomarker, cancer context, and literature scope. The system builds a task-specific knowledge graph, generates, validates candidate drug combinations, and returns a ranked list with supporting evidence.

Combination	Mechanism	Evidence	PMCID(s)
Afatinib + Fulvestrant	HER1-2 inhibition + estrogen-receptor modulation	Explicit KG paths	PMC6025235
Osimertinib + Palbociclib	EGFR resistance reversal + CDK4/6 inhibition	Embedding-inferred	PMC7471056
Lapatinib + Everolimus	HER1-2/neu receptor inhibition + mTOR inhibition	Explicit KG paths	PMC4072025
Dabrafenib + Trametinib	BRAF-MEK pathway inhibition	Embedding-inferred	PMC12992847, PMC7079252

TABLE I : Example hypotheses returned for the EGFR breast-cancer scenario. These are research candidates, not treatment recommendations.

Fig. 2 : Mean blinded-judge scores across seven biomarker-cancer scenarios for six evaluation dimensions. Llama-3.1-8B-Instruct is included as a matched zero-shot baseline for CoDHy. Error bars denote standard deviation across eligible judges; baseline generators abstain from judging their own outputs. Higher scores are better.

Fig. 3 : Blinded pairwise preferences for CoDHy versus each zer-shot baseline, aggregated across judges and scenarios. Percentages are reported from CoDHy’s perspective.

Explore similar work

HypoEvolve: Genetic Algorithms Enable Multi-Agent LLMs to Discover Scientific Hypotheses
Sep 14, 2026Jieyuan Liu, Mengzhou Hu, Jefferson Chen +10Hypothesis-Driven Model Expansion Multi-Agent Evolution
University of California San Diego · 2Texas A&M University · 3Carnegie Mellon University +1
Accelerating scientific discovery with Co-Scientist
Feb 26, 2025Juraj Gottweis, Wei-Hung Weng, Alexander Daryin +48Scientific Agents Scientific Discovery
BioInsight: Multi-Agent Orchestration for Interactive Biomedical Knowledge Discovery
Jun 19, 2026Jieyi Wang, Bingxuan Li, Nanyi Jiang +9Biomedical Knowledge Graph Multi-Agent Orchestration
1Peking University · University of Illinois at Urbana-Champaign · University of Pennsylvania +2

Sep 14, 2026, cs.CLJ/K move · Enter open · S save

HypoEvolve: Genetic Algorithms Enable Multi-Agent LLMs to Discover Scientific Hypotheses

Jieyuan Liu, Mengzhou Hu, Jefferson Chen, JungHo Kong+9

University of California San Diego · 2Texas A&M University · 3Carnegie Mellon University+1

Hypothesis-Driven Model Expansion Multi-Agent Evolution Scientific Agents Hypothesis Multi-Agent Large Language Model Systems Alphaevolve+3

Open paper AskPDF arXiv

Scientific agents contribute to hypothesis discovery by synthesizing evidence, assessing proposals, and developing new explanations. Recent systems combine scientific agents with evolutionary search through critique, comparison, and revision. However, how different forms of agent collaboration affect hypothesis quality remains an open question. Answering this question requires separating the effects of agents' scientific capabilities from those of their collaboration. A framework must therefore preserve agents' scientific roles and support rules for combining, revising, and retaining hypotheses. Building on this view, we introduce HypoEvolve, which makes collaboration explicit through successive updates to a hypothesis population. Specifically, we propose a generational genetic algorithm to coordinate specialized large language model (LLM) agents that integrate mechanistic arguments, reconsider assumptions, and assess evidence and testability. Each generation specifies how scientific judgments and new proposals reshape the population, making collaboration effects on hypothesis quality directly testable. Moreover, we design our evaluation around scientifically meaningful hypotheses that explain how a proposed intervention could work. Drug repurposing links these explanations to target-level biological claims assessed against external evidence. Specifically, we adapt DepMap and Open Targets into complementary external measures grounded in experimental, genetic, and clinical evidence. Across 34 cancer types, HypoEvolve achieves the highest scores against six baselines on both measures. DepMap selectivity reaches 0.171, versus 0.115 for the strongest baseline. Gains over single-pass generation also generalize to held-out cancer types. HypoEvolve advances a vision of autonomous science in which AI research teams achieve a capacity for discovery beyond that of individual models.

Abstract

Explore similar work

HypoEvolve: Genetic Algorithms Enable Multi-Agent LLMs to Discover Scientific Hypotheses

Accelerating scientific discovery with Co-Scientist

BioInsight: Multi-Agent Orchestration for Interactive Biomedical Knowledge Discovery