Rise of the Robots

Abstract

High‐throughput experimentation (HTE) is a growing, enabling technology that allows the execution of large, parallel sets of experiments. Often, automation is required to dose compounds on milligram to sub‐milligram scale, to run many parallel reactions, and to analyse large datasets. Unique approaches to screen design, implementation, and analysis are required, distinct from traditional synthetic organic chemistry. The discipline also presents a profitable opportunity for individual scientists to learn about and explore fields adjacent to chemistry, including data science, robotics and equipment engineering, and computer programming. This perspective presents the author's viewpoints on the field of HTE, its implementation within a chemistry career, and the automated future of organic chemistry technology.

Improving Data and Prediction Quality of High-Throughput Perovskite Synthesis with Model Fusion

“…These data were used to train four binary classifier models, where 1 ("positive") is the production of a large, high-quality single crystalline product, and 0 ("negative") is any other In this paper, our model fusion combines four single models in pairs (6), triples (4), and quadruple (1). Each of these 11 combined models is then considered using both score and rank combinations.…”

“…A model whose RSC function graph is a hypothetical diagonal line from point (0, 1.0) to point (1878, 0), corresponds to a simple linear relationship between score and rank. In Figure 1, the RSC function graph of the model WLC (C) model, which is closest to the diagonal line DL, assigns a score in [0, 1.0] to a rank in [1,1878] in proportionally decreasing order. An RSC function above this hypothetical diagonal line, such as the first 300 data items predicted by the GBT model (red), corresponds to assigning higher scores than to the corresponding ranks.…”

“…completion, the resultant crystals are scored by visual inspection into four outcome classes:(1) clear solution without any solid; (2) fine powder; (3) small crystallites (average crystal dimension < 0.1 mm); and (4) large (> 0.1 mm) crystals suitable for structure determination by single crystal X-ray diffraction. Of these, outcome class "4" corresponds to "positive" in our binary classification machine learning task.…”

Enabling Technology for Supramolecular Chemistry

“…High-throughput screening (HTS) enables reactions or processes to be rapidly carried out in parallel, potentially running thousands of samples simultaneously ( McNally et al, 2011 ). Whilst HTS may have been previously affected by a bottleneck in slower dispensing, and analytical or data-processing steps, its growing success has been enabled by the development of automated experimentation platforms and technologies such as liquid and solid handling devices, robotics, autosamplers, and data-analyst software or scripts ( Carson, 2020 ). It has been widely adopted within the pharmaceutical industry ( Mennen et al, 2019 ) and is now finding its use in many other chemical fields.…”

Small data materials design with machine learning: When the average model knows best

“…3,9,[19][20][21][22][23][24][25] In general, these achievements are rooted in the access to suitable large datasets, both theoretically and experimentally. [14][15][16][26][27][28] However, even though such big datasets (and access to them) are becoming common place, 3,27,[29][30][31][32] they do not represent the datasets most materials researchers work with on a day-to-day basis. Within general experimental material research projects, researchers generally produce no more than a hand full of data points (c.q.…”