Scalable and Cost-Efficient de Novo Template-Based Molecular Generation

2025/06/10 by Piotr Gaiński, Gaiński, Piotr, Oussama Boussif +13 · 1 voice · 2 citations
Biochemistry, Genetics and Molecular Biology · Computer Science · #Artificial Intelligence (cs.AI) #Biomolecules (q-bio.BM) #FOS: Biological sciences #FOS: Computer and information sciences #Machine Learning (cs.LG) #cs.AI #cs.LG #q-bio.BM

paper · pdf · doi:10.48550/arxiv.2506.19865

arxiv published 2025/06/10 · arxiv updated 2025/11/04

Abstract

Template-based molecular generation offers a promising avenue for drug design by ensuring generated compounds are synthetically accessible through predefined reaction templates and building blocks. In this work, we tackle three core challenges in template-based GFlowNets: (1) minimizing synthesis cost, (2) scaling to large building block libraries, and (3) effectively utilizing small fragment sets. We propose Recursive Cost Guidance, a backward policy framework that employs auxiliary machine learning models to approximate synthesis cost and viability. This guidance steers generation toward low-cost synthesis pathways, significantly enhancing cost-efficiency, molecular diversity, and quality, especially when paired with an Exploitation Penalty that balances the trade-off between exploration and exploitation. To enhance performance in smaller building block libraries, we develop a Dynamic Library mechanism that reuses intermediate high-reward states to construct full synthesis trees. Our approach establishes state-of-the-art results in template-based molecular generation.

Citations

• Action abstractions for amortized sampling
• Generative Flows on Synthetic Pathway for Drug Design
• Directly Optimizing for Synthesizability in Generative Molecular Design using Retrosynthesis Models
• RGFN: Synthesizable Molecular Generation Using GFlowNets
• Pessimistic Backward Policy for GFlowNets
• Generative Active Learning for the Search of Small-molecule Protein Binders
• Mixed Continuous and Categorical Flow Matching for 3D De Novo Molecule Generation
• Equivariant Flow Matching with Hybrid Probability Transport
• Towards Understanding and Improving GFlowNet Training
• Multi-Objective GFlowNets
• GFlowNets and variational inference
• Equivariant Diffusion for Molecule Generation in 3D
• Biological Sequence Design with GFlowNets
• Trajectory balance: Improved credit assignment in GFlowNets
• Amortized Tree Generation for Bottom-up Synthesis Planning and Synthesizable Molecular Design
• Flow Network based Generative Models for Non-Iterative Diverse Candidate Generation
• Therapeutics Data Commons: Machine Learning Datasets and Tasks for Drug Discovery and Development
• Learning To Navigate The Synthetically Accessible Chemical Space Using Reinforcement Learning
• Graph Transformer Networks
• Multi-Objective De Novo Drug Design with Conditional Graph Generative Model
• Count-Based Exploration with Neural Density Models
• #Exploration: A Study of Count-Based Exploration for Deep Reinforcement Learning
• Gaussian Error Linear Units (GELUs)
• Unifying Count-Based Exploration and Intrinsic Motivation
• Between MDPs and semi-MDPs: A framework for temporal abstraction in reinforcement learning

Cited by

• Diffusion Models at the Drug Discovery Frontier: A Review on Generating Small Molecules versus Therapeutic Peptides
• SynCoGen: Synthesizable 3D Molecule Generation via Joint Reaction and Coordinate Modeling

Discussions

• Gai\'nski, Boussif, Rekesh, Shevchuk, Parviz, Tyers, Batey, Koziarski: Scalable and Cost-Efficient de Novo Template-Based Molecular Generation https://arxiv.org/abs/2506.19865 https://arxiv.org/pdf/25 [bsky, 0 points, 0 comments]

Related

• Amortized Tree Generation for Bottom-up Synthesis Planning and Synthesizable Molecular Design
• SynFlowNet: Design of Diverse and Novel Molecules with Synthesis Constraints
• Generative Flows on Synthetic Pathway for Drug Design
• The Synthesizability of Molecules Proposed by Generative Models
• Learning To Navigate The Synthetically Accessible Chemical Space Using Reinforcement Learning