Transformer Performance for Chemical Reactions: Analysis of Different Predictive and Evaluation Scenarios

RLSynC: Offline-Online Reinforcement Learning for Synthon Completion

Retrosynthesis is the process of determining the set of reactant molecules that can react to form a desired product. Semitemplate-based retrosynthesis methods, which imitate the reverse logic of synthesis reactions, first predict the reaction centers in the products and then complete the resulting synthons back into reactants. We develop a new offline-online reinforcement learning method RLSynC for synthon completion in semitemplate-based methods. RLSynC assigns one agent to each synthon, all of which complete the synthons by conducting actions step by step in a synchronized fashion. RLSynC learns the policy from both offline training episodes and online interactions, which allows RLSynC to explore new reaction spaces. RLSynC uses a standalone forward synthesis model to evaluate the likelihood of the predicted reactants in synthesizing a product and thus guides the action search. Our results demonstrate that RLSynC can outperform state-of-the-art synthon completion methods with improvements as high as 14.9%, highlighting its potential in synthesis planning.

Artificial Intelligence Methods and Models for Retro-Biosynthesis: A Scoping Review

Retrosynthesis aims to efficiently plan the synthesis of desirable chemicals by strategically breaking down molecules into readily available building block compounds. Having a long history in chemistry, retro-biosynthesis has also been used in the fields of biocatalysis and synthetic biology. Artificial intelligence (AI) is driving us toward new frontiers in synthesis planning and the exploration of chemical spaces, arriving at an opportune moment for promoting bioproduction that would better align with green chemistry, enhancing environmental practices. In this review, we summarize the recent advancements in the application of AI methods and models for retrosynthetic and retro-biosynthetic pathway design. These techniques can be based either on reaction templates or generative models and require scoring functions and planning strategies to navigate through the retrosynthetic graph of possibilities. We finally discuss limitations and promising research directions in this field.

Deductive machine learning models for product identification

Deductive solution strategies are required in prediction scenarios that are under determined, when contradictory information is available, or more generally wherever one-to-many non-functional mappings occur. In contrast, most contemporary machine learning (ML) in the chemical sciences is inductive learning from example, with a fixed set of features. Chemical workflows are replete with situations requiring deduction, including many aspects of lab automation and spectral interpretation. Here, a general strategy is described for designing and training machine learning models capable of deduction that consists of combining individual inductive models into a larger deductive network. The training and testing of these models is demonstrated on the task of deducing reaction products from a mixture of spectral sources. The resulting models can distinguish between intended and unintended reaction outcomes and identify starting material based on a mixture of spectral sources. The models also perform well on tasks that they were not directly trained on, like performing structural inference using real rather than simulated spectral inputs, predicting minor products from named organic chemistry reactions, identifying reagents and isomers as plausible impurities, and handling missing or conflicting information. A new dataset of 1 124 043 simulated spectra that were generated to train these models is also distributed with this work. These findings demonstrate that deductive bottlenecks for chemical problems are not fundamentally insuperable for ML models.

Application of Transformers in Cheminformatics

By accelerating time-consuming processes with high efficiency, computing has become an essential part of many modern chemical pipelines. Machine learning is a class of computing methods that can discover patterns within chemical data and utilize this knowledge for a wide variety of downstream tasks, such as property prediction or substance generation. The complex and diverse chemical space requires complex machine learning architectures with great learning power. Recently, learning models based on transformer architectures have revolutionized multiple domains of machine learning, including natural language processing and computer vision. Naturally, there have been ongoing endeavors in adopting these techniques to the chemical domain, resulting in a surge of publications within a short period. The diversity of chemical structures, use cases, and learning models necessitate a comprehensive summarization of existing works. In this paper, we review recent innovations in adapting transformers to solve learning problems in chemistry. Because chemical data is diverse and complex, we structure our discussion based on chemical representations. Specifically, we highlight the strengths and weaknesses of each representation, the current progress of adapting transformer architectures, and future directions.

Large language models for science and medicine

Large language models (LLMs) are a type of machine learning model that learn statistical patterns over text, such as predicting the next words in a sequence of text. Both general purpose and task-specific LLMs have demonstrated potential across diverse applications. Science and medicine have many data types that are highly suitable for LLMs, such as scientific texts (publications, patents and textbooks), electronic medical records, large databases of DNA and protein sequences and chemical compounds. Carefully validated systems that can understand and reason across all these modalities may maximize benefits. Despite the inevitable limitations and caveats of any new technology and some uncertainties specific to LLMs, LLMs have the potential to be transformative in science and medicine.

Leveraging Language Model Multitasking To Predict C-H Borylation Selectivity

C-H borylation is a high-value transformation in the synthesis of lead candidates for the pharmaceutical industry because a wide array of downstream coupling reactions is available. However, predicting its regioselectivity, especially in drug-like molecules that may contain multiple heterocycles, is not a trivial task. Using a data set of borylation reactions from Reaxys, we explored how a language model originally trained on USPTO_500_MT, a broad-scope set of patent data, can be used to predict the C-H borylation reaction product in different modes: product generation and site reactivity classification. Our fine-tuned T5Chem multitask language model can generate the correct product in 79% of cases. It can also classify the reactive aromatic C-H bonds with 95% accuracy and 88% positive predictive value, exceeding purpose-developed graph-based neural networks.

Do Chemformers Dream of Organic Matter? Evaluating a Transformer Model for Multistep Retrosynthesis

Synthesis planning of new pharmaceutical compounds is a well-known bottleneck in modern drug design. Template-free methods, such as transformers, have recently been proposed as an alternative to template-based methods for single-step retrosynthetic predictions. Here, we trained and evaluated a transformer model, called the Chemformer, for retrosynthesis predictions within drug discovery. The proprietary data set used for training comprised ∼18 M reactions from literature, patents, and electronic lab notebooks. Chemformer was evaluated for the purpose of both single-step and multistep retrosynthesis. We found that the single-step performance of Chemformer was especially good on reaction classes common in drug discovery, with most reaction classes showing a top-10 round-trip accuracy above 0.97. Moreover, Chemformer reached a higher round-trip accuracy compared to that of a template-based model. By analyzing multistep retrosynthesis experiments, we observed that Chemformer found synthetic routes, leading to commercial starting materials for 95% of the target compounds, an increase of more than 20% compared to the template-based model on a proprietary compound data set. In addition to this, we discovered that Chemformer suggested novel disconnections corresponding to reaction templates, which are not included in the template-based model. These findings were further supported by a publicly available ChEMBL compound data set. The conclusions drawn from this work allow for the design of a synthesis planning tool where template-based and template-free models work in harmony to optimize retrosynthetic recommendations.

SynRoute: A Retrosynthetic Planning Software

Computer-assisted synthetic planning has seen major advancements that stem from the availability of large reaction databases and artificial intelligence methodologies. SynRoute is a new retrosynthetic planning software tool that uses a relatively small number of general reaction templates, currently 263, along with a literature-based reaction database to find short, practical synthetic routes for target compounds. For each reaction template, a machine learning classifier is trained using data from the Pistachio reaction database to predict whether new computer-generated reactions based on the template are likely to work experimentally in the laboratory. This reaction generation methodology is used together with a vectorized Dijkstra-like search of top-scoring routes organized by synthetic strategies for easy browsing by a synthetic chemist. SynRoute was able to find routes for an average of 83% of compounds based on selection of random subsets of drug-like compounds from the ChEMBL database. Laboratory evaluation of 12 routes produced by SynRoute, to synthesize compounds not from the previous random subsets, demonstrated the ability to produce feasible overall synthetic strategies for all compounds evaluated.