Self-driving labs and automation software for chemistry and materials science

Electrochemical research often requires stringent combinations of experimental parameters that are demanding to manually locate. Here the authors report an autonomous electrochemical platform that implements an adaptive, closed-loop workflow for mechanistic investigation of molecular electrochemistry.

Autonomous exploration of materials design space is hindered by its high dimensionality and the scarcity of data. In this work, we present AlphaFlow, a self-driven lab guided by reinforcement learning that enables accelerated discovery and optimization of multi-step chemistries.

Spirocyclic tetrahydronaphthyridines (THNs) are valuable structural motifs in medicinal chemistry, but the modular and scalable synthesis of this specific motif remains challenging. Here, the authors develop an automated and continuous flow synthesis of 1,8-THN and 1,6-THN analogues based on photoredox-catalysed hydroaminoalkylation, demonstrating the concise synthesis of the spirocyclic THN core of Pfizer’s MC4R antagonist PF-07258669.

Late-stage functionalization of drug molecules can tune their properties without the need for entirely new syntheses, however, predicting reactivity and planning synthesis for late-stage C-H activation remains challenging. Here, the authors develop a reaction screening approach combining high-throughput experimentation with computational graph neural networks to identify suitable substrates that can be used for late-stage C-H alkylation via Minisci-type chemistry.

Ozone-induced dissociation (OzID) coupled with ion mobility spectrometry-mass spectrometry (IMS-MS) provides the capacity for in-depth structural elucidation of lipids with isomer separation and confident assignment of double bond positions, however, OzID data analysis remains very challenging. Here, the authors develop a Python tool, LipidOz, for the automated determination of lipid double bond locations from complex LC-OzID-IMS-MS data, with a combination of traditional automation and deep learning approaches.

Chemical structures are typically published as nonmachine-readable images in scientific literature. Here, the authors present DECIMER.ai, an open platform for translating chemical structures in publications into machine-readable representations.

Materials language and processing with large language models provide an automated approach for text classification. Here, a generative pretrained transformer (GPT) approach is reported to provide a simple architecture for text classification, including identifying incorrectly annotated data and for manual labelling.

Predicting properties at the interface of materials is crucial for advanced materials design. Here, the authors introduce a high-throughput computational framework, InterMatch, for predicting several properties of an interface by using the databases of individual bulk materials.

Global challenges demand global solutions. Here, the authors show a distributed self-driving lab architecture in The World Avatar, linking robots in Cambridge and Singapore for asynchronous multi-objective reaction optimisation.