Open-source Polymer Generative Pipeline
Journal:
arXiv
Published Date:
Nov 29, 2024
Abstract
Polymers play a crucial role in the development of engineering materials,
with applications ranging from mechanical to biomedical fields. However, the
limited polymerization processes constrain the variety of organic building
blocks that can be experimentally tested. We propose an open-source
computational generative pipeline that integrates neural-network-based
discriminators, generators, and query-based filtration mechanisms to overcome
this limitation and generate hypothetical polymers. The pipeline targets
properties, such as ionization potential (IP), by aligning various
representational formats to generate hypothetical polymer candidates. The
discriminators demonstrate improvements over state-of-the-art models due to
optimized architecture, while the generators produce novel polymers tailored to
the desired property range. We conducted extensive evaluations to assess the
generative performance of the pipeline components, focusing on the polymers'
ionization potential (IP). The developed pipeline is integrated into the
DeepChem framework, enhancing its accessibility and compatibility for various
polymer generation studies.
