Within the last decade advances in computer hardware and user-friendly software libraries have led to a surge in popularity for computer-assisted data analysis techniques. Notably, machine learning (ML) and artificial intelligence (AI) techniques have been applied to large and diverse datasets, hitherto intractable to simple modelling schemes. The Cambridge Structural Database (CSD) is a particularly appealing dataset on which to deploy these emerging tools due to its position as the world’s largest repository for experimentally determined organic and metal-organic crystal structures, with nearly 1.1 million entries as of January 2021. The structures contained within the CSD reflect a broad range of diverse chemistries from over 400,000 authors who have deposited to it over the last fifty-five years, resulting in a unique, rich dataset from which to extract knowledge.<br/><br/>This talk will highlight some of the ways researchers, from within and outside the CCDC, have utilized the diverse structural data in the CSD to gain new insights into chemistry, biochemistry, and materials science by applying ML and AI techniques. Beyond simply sourcing large amounts of data, researchers also face the challenge of ensuring the quality and integrity of the data they input into their models. We will therefore also examine the work undertaken to ensure the quality of data found within resources such as the CSD, so that they can be used as a robust foundation for uncovering meaningful knowledge and insights.

Learning from a million structures: Data mining, machine learning, AI and the Cambridge structural database

ACS Spring Meeting 2021

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.