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

2025-08-02

AI Summary

AI-driven computer-aided synthetic planning (CASP) suffers from high inference latency, hindering high-throughput synthesizability screening in de novo drug design. To address this, we propose a low-latency, multi-step retrosynthetic planning acceleration framework tailored for Transformer architectures. Our method innovatively integrates speculative beam search with a scalable Medusa draft strategy, enabling parallelized, low-latency inference atop a SMILES-to-SMILES single-step predictor. Experiments under strict time budgets of several seconds demonstrate that our approach improves molecular solution coverage by 26%–86% over baseline methods, substantially enhancing retrosynthetic throughput. This work represents the first systematic application of speculative decoding to CASP, establishing an efficient and practical pathway for real-time, high-throughput assessment of drug molecule synthesizability.

Technology Category

• Search and Optimization: Sampling/Simulation-based Search
• Planning, Routing, and Scheduling: Planning with Language Models
• Multiagent Systems: Multiagent Planning

Application Category

• Search and Retrieval-Augmented AI: Agentic search
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

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.

Problem

• Reducing high latency in AI-based synthesis planning systems
• Accelerating multi-step retrosynthesis with transformer models
• Meeting strict latency needs for high-throughput drug screening

Innovation

• Transformer neural network for retrosynthetic planning
• Speculative beam search to reduce latency
• Scalable Medusa drafting strategy