Fast and scalable retrosynthetic planning with a transformer neural network and speculative beam search

TL;DR

This paper introduces a fast, scalable retrosynthetic planning method using transformer neural networks and speculative beam search, significantly reducing latency and increasing molecule synthesis predictions within strict time constraints.

Contribution

The authors develop a novel speculative beam search combined with Medusa drafting strategy to accelerate multi-step retrosynthesis planning in AI systems.

Findings

Achieves 26-86% increase in molecules solved within time limits.

Reduces latency of transformer-based retrosynthesis models.

Enhances suitability of AI CASP for high-throughput screening.

Abstract

AI-based computer-aided synthesis planning (CASP) systems are in demand as components of AI-driven drug discovery workflows. However, the high latency of such CASP systems limits their utility for high-throughput synthesizability screening in de novo drug design. We propose a method for accelerating multi-step synthesis planning systems that rely on SMILES-to-SMILES transformers as single-step retrosynthesis models. Our approach reduces the latency of SMILES-to-SMILES transformers powering multi-step synthesis planning in AiZynthFinder through speculative beam search combined with a scalable drafting strategy called Medusa. Replacing standard beam search with our approach allows the CASP system to solve 26\% to 86\% more molecules under the same time constraints of several seconds. Our method brings AI-based CASP systems closer to meeting the strict latency requirements of high-throughput synthesizability screening and improving general user experience.