Chemical knowledge-informed framework for privacy-aware retrosynthesis learning

TL;DR

This paper introduces CKIF, a privacy-preserving framework for retrosynthesis learning that enables multiple chemical organizations to collaboratively train models without sharing proprietary data, leveraging chemical knowledge for model aggregation.

Contribution

The paper presents a novel distributed learning framework that uses chemical knowledge to ensure privacy while improving retrosynthesis prediction accuracy.

Findings

CKIF outperforms baseline methods on various reaction datasets.

Chemical properties guide adaptive model aggregation effectively.

Framework maintains data confidentiality across organizations.

Abstract

Chemical reaction data is a pivotal asset, driving advances in competitive fields such as pharmaceuticals, materials science, and industrial chemistry. Its proprietary nature renders it sensitive, as it often includes confidential insights and competitive advantages organizations strive to protect. However, in contrast to this need for confidentiality, the current standard training paradigm for machine learning-based retrosynthesis gathers reaction data from multiple sources into one single edge to train prediction models. This paradigm poses considerable privacy risks as it necessitates broad data availability across organizational boundaries and frequent data transmission between entities, potentially exposing proprietary information to unauthorized access or interception during storage and transfer. In the present study, we introduce the chemical knowledge-informed framework (CKIF), a privacy-preserving approach for learning retrosynthesis models. CKIF enables distributed training across multiple chemical organizations without compromising the confidentiality of proprietary reaction data. Instead of gathering raw reaction data, CKIF learns retrosynthesis models through iterative, chemical knowledge-informed aggregation of model parameters. In particular, the chemical properties of predicted reactants are leveraged to quantitatively assess the observable behaviors of individual models, which in turn determines the adaptive weights used for model aggregation. On a variety of reaction datasets, CKIF outperforms several strong baselines by a clear margin.