Large Language Models for Controllable Multi-property Multi-objective Molecule Optimization

TL;DR

This paper introduces GeLLMO-Cs, instruction-tuned large language models designed for multi-property, multi-objective molecule optimization, demonstrating superior performance and generalization in drug design tasks.

Contribution

The paper presents C-MuMOInstruct, a novel instruction-tuning dataset for multi-property molecule optimization, and develops GeLLMO-Cs models that excel in property-specific, multi-objective molecule optimization tasks.

Findings

GeLLMO-Cs outperform baselines with up to 126% higher success rate.

Models exhibit strong 0-shot generalization to new tasks and instructions.

Effective for real-world drug design with diverse property objectives.

Abstract

In real-world drug design, molecule optimization requires selectively improving multiple molecular properties up to pharmaceutically relevant levels, while maintaining others that already meet such criteria. However, existing computational approaches and instruction-tuned LLMs fail to capture such nuanced property-specific objectives, limiting their practical applicability. To address this, we introduce C-MuMOInstruct, the first instruction-tuning dataset focused on multi-property optimization with explicit, property-specific objectives. Leveraging C-MuMOInstruct, we develop GeLLMO-Cs, a series of instruction-tuned LLMs that can perform targeted property-specific optimization. Our experiments across 5 in-distribution and 5 out-of-distribution tasks show that GeLLMO-Cs consistently outperform strong baselines, achieving up to 126% higher success rate. Notably, GeLLMO-Cs exhibit impressive 0-shot generalization to novel optimization tasks and unseen instructions. This offers a step toward a foundational LLM to support realistic, diverse optimizations with property-specific objectives. C-MuMOInstruct and code are accessible through https://github.com/ninglab/GeLLMO-C.