Quantum Chemical Data Generation as Fill-In for Reliability Enhancement of Machine-Learning Reaction and Retrosynthesis Planning

Abstract

Data-driven synthesis planning has seen remarkable successes in recent years by virtue of modern approaches of artificial intelligence that efficiently exploit vast databases with experimental data on chemical reactions. However, this success story is intimately connected to the availability of existing experimental data. It may well occur in retrosynthetic and synthesis design tasks that predictions in individual steps of a reaction cascade are affected by large uncertainties. In such cases, it will, in general, not be easily possible to provide missing data from autonomously conducted experiments on demand. However, first-principles calculations can, in principle, provide missing data to enhance the confidence of an individual prediction or for model retraining. Here, we demonstrate the feasibility of such an ansatz and examine resource requirements for conducting autonomous first-principles calculations on demand.

Pathfinder─Navigating and Analyzing Chemical Reaction Networks with an Efficient Graph-Based Approach

“…or hazardous reactions in silico to determine, if a targeted product is formed at all, what kind of (hazardous) side products might emerge during the course of a reaction and what is their ratio compared to the targeted product without relying on kinetic modeling (cf. ref ( 31 )).…”

Quantum chemical data generation as fill-in for reliability enhancement of machine-learning reaction and retrosynthesis planning

“…The data underlying the results presented in this article is available on Zenodo. 84 This exploration has been carried out with Chemoton 2.1.0 in conjunction with Puffin 1.1.0. Note that this implies the following versions of the entire SCINE Chemoton exploration stack:…”

Accelerating Reaction Network Explorations with Automated Reaction Template Extraction and Application

“…[14 -19 ] We have recently consolidated experiment-based data learned by a SMILES [20 ]based transformer [21 ] with ab inito calculations and presented some of the challenges. [18 ] Experimental data of syntheses abstracts multi-step processes while quantum chemical studies fully resolve mechanistic understanding at the level of elementary steps and atoms. This leaves a gap in the knowledge transfer chain that has to be filled with expensive, brute-force exploration steps if no other methods are available.…”