Understanding structural adaptability: a reactant informatics approach to experiment design

Abstract

The structural and electronic adaptability of a vanadium selenite framework is determined using cheminformatics data and machine learning algorithms.

Autonomous Discovery in the Chemical Sciences Part I: Progress

“…used a combination of 3955 reaction successes and failures from laboratory notebooks to train a support vector machine (SVM) model to predict outcomes for the crystallization of vanadium selenites [184] (Figure 9). Recasting the model as a decision tree led to correlations that reflected expert intuition, which arguably contributed to the synthesis of five previously unseen compounds [185] . A similar study applied a much smaller dataset of 54 conditions to predict whether a process would produce atomically precise gold nanocrystals, using a siamese neural network architecture to relate proposed conditions to precedents [186] .…”

“…Recasting the model as adecision tree led to correlations that reflected expert intuition, which arguably contributed to the synthesis of five previously unseen compounds. [185] As imilar study applied am uch smaller dataset of 54 conditions to predict whether ap rocess would produce atomically precise gold nanocrystals,u sing as iamese neural network architecture to relate proposed conditions to precedents. [186] Forl arger-scale analyses,t he literature serves as an unstructured data source of inorganic reactions and has been used to populate as tructured database of synthesis conditions and outcomes via natural language processing of over 640 000 manuscripts; [187] virtual screening and synthesis planning pipelines have been built on top of such data to guide the experimental realization of computationally proposed materials.…”

Twenty years of exceptional success: The molecular education and research consortium in undergraduate computational chemistry (MERCURY)

“…The faculty published 2 peer‐reviewed books, [ 238,239 ] 9 peer‐reviewed book chapters, [ 240–248 ] and 115 peer‐reviewed research papers. [ 249–363 ] This comes to 1.6 peer‐reviewed products/faculty/year during the 3‐year grant period, which is 3.2 times the rate of publication for natural science faculty at PUIs. [ 46 ]…”

“…Joshua Schrier's research: The Schrier group uses simulations and machine learning to understand and design organic‐inorganic hybrid materials for energy and environmental problems. In particular, they are interested in understanding the role of noncovalent interactions in controlling inorganic topologies, [ 343 ] the role of reaction conditions and reagent properties on reaction outcome, [ 330 ] and how human biases in experimental decision making hinder progress in machine learning using those datasets. [ 344 ] Current computational work will contribute to understanding molecular interactions in the formation of precursor solutions and electrolyte solution properties.…”

Autonome Entdeckung in den chemischen Wissenschaften, Teil I: Fortschritt

“…Dadurch konnten fünf zuvor unbekannte Verbindungen synthetisiert werden. [185] In einer ähnlichen Studie wurde mit einem viel kleineren Datensatz aus 54 Bedingungen vorhergesagt, ob Goldatome zu Nanokristallen kristallisieren würden, wobei eine siamesische neuronale Netzwerkarchitektur die vorgeschlagenen Bedingungen mit bekannten Bedingungen abglich. [186] Fürgrçßerskalige Analysen bietet die Literatur eine unstrukturierte Datenquelle mit anorganischen Reaktionen, und durch natürliche Sprachverarbeitung von 640 000 Manuskripten wurde eine strukturierte Datenbank mit Synthesebedingungen und Ausbeuten gefüllt; [187] anhand der Daten wurden virtuelle Screening-und Planungspipelines aufgebaut, als Anleitung fürd ie Umsetzung des Computervorschlags zur experimentellen Herstellung des Materials.…”