Efficient and Programmable Exploration of Synthesizable Chemical Space

2025-11-29

AI Summary

This work addresses the challenge of efficient, programmable molecular discovery within synthetically accessible chemical space. We propose PrexSyn, the first generative framework supporting both multi-attribute logical programming (e.g., “active ∧ low-toxicity ∧ synthesizable”) and iterative optimization over black-box objective functions. Methodologically, PrexSyn employs a decoder-only Transformer architecture trained on a billion-scale dataset of synthesis-pathway–molecular-property pairs, generated via a high-throughput C++ synthesis engine. Compared to state-of-the-art approaches, PrexSyn achieves superior synthetic space coverage, sampling efficiency, and inference speed. It attains high reconstruction accuracy and—critically—establishes the first unified paradigm for molecule design that is attribute-programmable, iteration-optimizable, and generation-verifiable.

Technology Category

Natural Language Processing: Code Generation / Program Synthesis from Natural LanguageSearch and Optimization: Mixed Discrete/Continuous SearchMachine Learning: Deep Generative Models & Autoencoders

Application Category

Search and Retrieval-Augmented AI: Large language models for searchSemantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMsWeb Mining and Content Analysis: Large pretrained models with web data

Abstract

The constrained nature of synthesizable chemical space poses a significant challenge for sampling molecules that are both synthetically accessible and possess desired properties. In this work, we present PrexSyn, an efficient and programmable model for molecular discovery within synthesizable chemical space. PrexSyn is based on a decoder-only transformer trained on a billion-scale datastream of synthesizable pathways paired with molecular properties, enabled by a real-time, high-throughput C++-based data generation engine. The large-scale training data allows PrexSyn to reconstruct the synthesizable chemical space nearly perfectly at a high inference speed and learn the association between properties and synthesizable molecules. Based on its learned property-pathway mappings, PrexSyn can generate synthesizable molecules that satisfy not only single-property conditions but also composite property queries joined by logical operators, thereby allowing users to ``program'' generation objectives. Moreover, by exploiting this property-based querying capability, PrexSyn can efficiently optimize molecules against black-box oracle functions via iterative query refinement, achieving higher sampling efficiency than even synthesis-agnostic baselines, making PrexSyn a powerful general-purpose molecular optimization tool. Overall, PrexSyn pushes the frontier of synthesizable molecular design by setting a new state of the art in synthesizable chemical space coverage, molecular sampling efficiency, and inference speed.

Problem

• Efficiently explores synthesizable chemical space for molecules
• Generates molecules meeting single or composite property queries
• Optimizes molecules via iterative refinement against black-box functions

Innovation

• Decoder-only transformer trained on billion-scale synthesizable pathway data
• Generates molecules via programmable composite property queries with logical operators
• Optimizes molecules via iterative query refinement against black-box oracle functions