RetroOOD: Understanding Out-of-Distribution Generalization in Retrosynthesis Prediction

Authors

  • Yemin Yu City University of Hong Kong Shanghai Institute for Advanced Study of Zhejiang University, China
  • Luotian Yuan Zhejiang University, China
  • Ying Wei Nanyang Technology University
  • Hanyu Gao Hong Kong University of Science and Technology
  • Fei Wu Shanghai Institute for Advanced Study of Zhejiang University, China Zhejiang University, China
  • Zhihua Wang Shanghai Institute for Advanced Study of Zhejiang University, China
  • Xinhai Ye Zhejiang University

DOI:

https://doi.org/10.1609/aaai.v38i1.27791

Keywords:

APP: Natural Sciences

Abstract

Machine learning-assisted retrosynthesis prediction models have been gaining widespread adoption, though their performances oftentimes degrade significantly when deployed in real-world applications embracing out-of-distribution (OOD) molecules or reactions. Despite steady progress on standard benchmarks, our understanding of existing retrosynthesis prediction models under the premise of distribution shifts remains stagnant. To this end, we first formally sort out two types of distribution shifts in retrosynthesis prediction and construct two groups of benchmark datasets. Next, through comprehensive experiments, we systematically compare state-of-the-art retrosynthesis prediction models on the two groups of benchmarks, revealing the limitations of previous in-distribution evaluation and re-examining the advantages of each model. More remarkably, we are motivated by the above empirical insights to propose two model-agnostic techniques that can improve the OOD generalization of arbitrary off-the-shelf retrosynthesis prediction algorithms. Our preliminary experiments show their high potential with an average performance improvement of 4.6%, and the established benchmarks serve as a foothold for further retrosynthesis prediction research towards OOD generalization.

Downloads

Published

2024-03-25

How to Cite

Yu, Y., Yuan, L., Wei, Y., Gao, H., Wu, F., Wang, Z., & Ye, X. (2024). RetroOOD: Understanding Out-of-Distribution Generalization in Retrosynthesis Prediction. Proceedings of the AAAI Conference on Artificial Intelligence, 38(1), 374-382. https://doi.org/10.1609/aaai.v38i1.27791

Issue

Section

AAAI Technical Track on Application Domains