Data Augmentation and Pretraining for Template-Based Retrosynthetic Prediction in Computer-Aided Synthesis Planning

Abstract

<jats:p>This work presents efforts to augment the performance of data-driven machine learning algorithms for reaction template recommendation used in computer-aided synthesis planning software. Often, machine learning models designed to perform the task of prioritizing reaction templates or molecular transformations are focused on reporting high accuracy metrics for the one-to-one mapping of product molecules in reaction databases to the template extracted from the recorded reaction. The available templates that get selected for inclusion in these machine learning models have been previously limited to those that appear frequently in the reaction databases and exclude potentially useful transformations. By augmenting open-access datasets of organic reactions with artificially calculated template applicability and pretraining a template relevance neural network on this augmented applicability dataset, we report an increase in the template applicability recall and an increase in the diversity of predicted precursors. The augmentation and pretraining effectively teaches the neural network an increased set of templates that could theoretically lead to successful reactions for a given target. Even on a small dataset of well curated reactions, the data augmentation and pretraining methods resulted in an increase in top-1 accuracy, especially for rare templates, indicating these strategies can be very useful for small datasets.</jats:p>

“Ring Breaker”: Neural Network Driven Synthesis Prediction of the Ring System Chemical Space

“…Furthermore, we propose that this methodology can be extended to other specialized domains within synthesis planning tasks where the data may be limited and domain specific knowledge (i.e., a specialist) is required. The methodology could also be extended to combine various data sources to increase domain specific coverage, in addition to data augmentation techniques published at the time of writing this manuscript …”

Atom‐to‐atom Mapping: A Benchmarking Study of Popular Mapping Algorithms and Consensus Strategies

“…Atom-to-atom mapping (AAM) [6,7] is a procedure that establishes a correspondence between the atoms of reactants and products. AAM allows to identify a reaction centre (RC) which, in turn, helps to prepare reaction templates used in an automatized forward/retrosynthesis planning, [8][9][10][11][12] as well as to perform reaction classification [13] and reaction searching. [14,15] Several publicly and commercially available AAM tools are currently available.…”

Efficient Data Undersampling for Rule-Based Retrosynthetic Planning

“…Second, the amount of data is much greater than the domain knowledge of individual researchers. Therefore, with the recent rapid progress of deep learning, the use of machine learning algorithms to learn the latent space of retrosynthetic reaction rules has become a very active research area [1][2][3][4][5][6][7][8][9][10][11]. A rule-based model is a machine learning algorithm that learns the reaction rules that correspond to the input target molecules.…”

“…The reaction center means changing parts (atom and bonding) before and after the chemical reaction. Hence, the strength of rule-based models [1][2][3][4] is that it is easy to identify the selected reaction rules for the target molecule compared with the molecular transformer based on sequence-tosequence models and the attention mechanism [5][6][7][8][9].…”