English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

An essential aspect for adequate predictions of chemical properties bymachine learning models is the database used for training them. However,studies that analyze how the content and structure of the databases used fortraining impact the prediction quality are scarce. In this work, we analyze andquantify the relationships learned by a machine learning model (Neural Network)trained on five different reference databases (QM9, PC9, ANI-1E, ANI-1 andANI-1x) to predict tautomerization energies from molecules in Tautobase. Forthis, characteristics such as the number of heavy atoms in a molecule, numberof atoms of a given element, bond composition, or initial geometry on thequality of the predictions are considered. The results indicate that trainingon a chemically diverse database is crucial for obtaining good results but alsothat conformational sampling can partly compensate for limited coverage ofchemical diversity. We explicitly demonstrate that when certain types of bondsneed to be covered in the target database (Tautobase) but are undersampled inthe reference databases the resulting predictions are poor. A quantitativemeasure for these deficiencies is the Kullback-Leibler divergence betweenreference and target distributions. Analysis of the results with a TreeMAPalgorithm provides deeper understanding of specific deficiencies in thereference data sets. Capitalizing on this information can be used to eitherimprove existing databases or to generate new databases of sufficient diversityfor a range of ML applications in chemistry.

Impact of the characteristics of quantum chemical databases on machine learning predictions of tautomerization energies