DrugCoding: A Scientific Programming Agent for Drug Discovery
AuthorsJunwei Hu, Zhuohao Lin, Kun Li, Hengnan Zhang
AffiliationsXDeNovo Labs, Shanghai, China
DrugCoding is an AI agent that plans and executes multi-step computational tasks to help automate drug discovery.
Key results
Complete multi-turn sessions in the benchmark.
Tool-call inputs across the benchmark's eight interfaces.
DrugCoding's average session-level score on TCDD.
Absolute percentage-point lead over the strongest coding-agent baseline.
DrugCoding's average score on the challenging TCDD subset.
Average Pass@1 improvement over the single-agent configuration.
What the paper found
DrugCoding is a scientific programming agent for coordinating drug-discovery workflows. It decomposes research requests into dependency-aware tasks, routes each task to an eligible specialist model, and combines role-specific agents with programmatic execution, result validation, and provenance tracking. Its supported capabilities include molecular generation, virtual screening, docking, ADMET prediction, and drug–target affinity prediction, using tools such as DiffDock, DrugCLIP, and ADMET-AI. On the multi-turn TCDD benchmark, which contains 1,250 sessions and 2,500 action-level inputs, DrugCoding achieved 97.35% average session-level Pass@1—9.24 percentage points above OpenAI Codex, its strongest coding-agent baseline. On the challenging subset, it scored 97.92% Pass@1, ahead of Anthropic’s Claude Science and Rosalind Workbench. The evaluation tests whether agents produce correctly ordered tool calls with exact required inputs, not whether the resulting scientific predictions are experimentally valid. An ablation found that multi-agent coordination raised average Pass@1 by 0.56 percentage points, though gains varied by language model. Failure analysis points to fewer input-preservation and structured-output errors as recurring strengths.
Original abstract
Drug discovery involves multiple interconnected research stages, including molecular design, property prediction, and virtual screening. These processes are characterized by complex workflows, interdependent computational tasks, and diverse specialized tools, placing substantial demands on the coordination of scientific models and computational resources. However, existing large language model (LLM)-based scientific agents still face challenges in autonomous planning of complex tasks, adaptive selection of specialized models, and reliable execution of multi-stage scientific computations, limiting their effectiveness in automating complex drug discovery workflows. To address these challenges, we propose DrugCoding, a scientific programming agent system that integrates scientific task planning, adaptive model routing, multi-agent collaboration, and programmatic tool execution. DrugCoding enables autonomous task planning and dynamic model adaptation for complex research objectives. It can flexibly invoke specialized scientific models, autonomously generate and execute code according to task requirements, and leverage multi-agent collaboration to support efficient execution and reliable validation of complex scientific tasks. To evaluate DrugCoding's performance in scientific tool invocation and multi-stage task execution, we conduct multi-turn tool-calling experiments across diverse language models and scientific tool interfaces. Experimental results demonstrate that DrugCoding achieves an average session-level Pass@1 of 97.35% on the full benchmark, outperforming the strongest baseline, Codex, by 9.24%. On the challenging subset, DrugCoding achieves 97.92%, exceeding the strongest baseline.
Read the original paperMore in AI for Science
Browse all 58 papers →Reliability by Architecture: Agentic Workflows for Validated Multi-Physics Simulation from Natural Language
Søren Taverniers, Eric Darve, Hamdi A. Tchelepi
An LLM-assisted workflow prepares and repairs multi-physics simulations, while deterministic checks help ensure the results match both the physics and the user’s request.
Sample-Efficient Control of Magnetoconvective Flows via Differentiable Simulation
Jannis Becktepe, Dmitry Krasnov, Nils Thuerey, Jörg Schumacher, Sebastian Peitz
A differentiable simulator helps researchers control magnetoconvective flows more efficiently and design better heat-transfer outcomes.
Blind Variation, Selective Retention: Improving Creative Idea Generation in AI Scientists
Guowei Xu, Jiaru Zou, Mete Erdogan, Haotian Ye, Pan Lu, James Zou
The paper finds that AI scientists may generate more creative ideas by first knowing less about a task, then using evidence to choose the strongest ones.