Mining Electronic Laboratory Notebooks: Analysis, Retrosynthesis, and Reaction Based Enumeration

Abstract

An approach to automatically analyze and use the knowledge contained in electronic laboratory notebooks (ELNs) has been developed. Reactions were reduced to their reactive center and converted to a string representation (SMIRKS) which formed the basis for reaction classification and in silico (retro-)synthesis. Of the SMIRKS that occurred at least five times, 98% successfully regenerated the original product. The extracted reaction rules (SMIRKS) and corresponding reactants span a virtual chemical space which showed a strong dependence on the size of the reactive center. Whereas relatively few robust reaction types were sufficient to describe a large part of all reactions, considerably more reaction rules were necessary to cover all reactions in the ELN. Furthermore, reaction sequences were extracted to identify frequent combinations and diversifying reaction steps. Based on the extracted knowledge a (retro-)synthesis tool was built allowing for de novo design of compounds which have a high chance of being synthetically accessible. In an example application of the de novo design tool, various feasible retrosynthetic routes to the query molecule were obtained. Reaction based enumeration along the top ranked route yielded a library of 29 920 compounds with diverse properties, 99.9% of which are novel in the sense that they are unknown to the public domain.

Modelling Chemical Reasoning to Predict and Invent Reactions

“…As the rule base, we extracted reaction rules algorithmically from the same dataset that was used to build the knowledge graph. Here, the rules contained the reaction centre and additionally the atoms neighbouring the reaction centre up to a depth of two bonds (see Section S1.5 in the Supporting Information) . This rule definition usually also grasps the necessary activating groups, for example, carbonyl groups in α‐alkylations.…”

Toward Design of Novel Materials for Organic Electronics

“…In the past, automatic retrosynthesis or reaction prediction (see Figure 5a) required information from databases and/or the manual encoding of chemical rules. Newer developments include the automated extraction of reaction rules, [143] new models for chemical reasoning, [144] heuristics aided methods, [145] and the use of machine learning. Some of the most widely used systems are ChemPlanner, [141] PathFinder, ICSynth, LHASA, CAMEO, SOPHIA, and EROS.…”

Alternative methods of processing bio-feedstocks in formulated consumer product design

“…In the past retrosynthetic analysis was limited by the requirement for an expert chemist to manually program reaction rules. Law et al (2009) addressed this problem by using the Beilstein Crossfire database to automatically incorporate all known reactions into the reaction rule generation, while (Christ et al, 2012) reports of efforts at Boehringer Ingelheim to mine electronic laboratory notebooks for reaction rules, extending their reach past the published chemistry. Huang et al (2011) has sought to address the issue that though a reaction might be feasible on paper it can be very difficult or inefficient to conduct in practice by introducing an accessibility factor allowing for a better differentiation of routes.…”