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

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Gaiński, Piotr, Boussif, Oussama, Rekesh, Andrei, Shevchuk, Dmytro, Parviz, Ali, Tyers, Mike, Batey, Robert A., Koziarski, Michał
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866908625629347840
author Gaiński, Piotr
Boussif, Oussama
Rekesh, Andrei
Shevchuk, Dmytro
Parviz, Ali
Tyers, Mike
Batey, Robert A.
Koziarski, Michał
author_facet Gaiński, Piotr
Boussif, Oussama
Rekesh, Andrei
Shevchuk, Dmytro
Parviz, Ali
Tyers, Mike
Batey, Robert A.
Koziarski, Michał
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2506_19865
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Scalable and Cost-Efficient de Novo Template-Based Molecular Generation
Gaiński, Piotr
Boussif, Oussama
Rekesh, Andrei
Shevchuk, Dmytro
Parviz, Ali
Tyers, Mike
Batey, Robert A.
Koziarski, Michał
Biomolecules
Artificial Intelligence
Machine Learning
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.
title Scalable and Cost-Efficient de Novo Template-Based Molecular Generation
topic Biomolecules
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2506.19865