A Precedent-Guided Co-Scientist for Side-Effect-Aware Drug Redesign

Yujin Kim, Charmgil Hong

Track: Track 3: AI Scientist Proposal Competition

Keywords: Drug Redesign, Agentic Workflow, Side-Effect Mitigation, Evidence-Grounded Reasoning

TL;DR: PRECEDE is a human-supervised LLM co-scientist that grounds side-effect-aware drug redesign in biomedical evidence and historical redesign precedents.

Abstract: We propose PRECEDE, a precedent-guided co-scientist for side-effect-aware drug redesign that revises a parent compound to mitigate a specified adverse effect while preserving therapeutic function. Rather than isolated molecular generation, PRECEDE frames redesign as evidence-grounded reasoning over drug--side effect associations, biomedical knowledge graphs, and structured precedents of prior safety-driven optimization, coordinated by an LLM orchestrator with explicit decision policies and human checkpoints. We position PRECEDE as a human-supervised AI-for-science workflow whose hypotheses remain auditable, falsifiable, and bounded by what prior pharmacology already supports.

Submission Number: 264