A Structure-Based Platform for Predicting Chemical Reactivity

Abstract

Although machine learning has a long-standing history in chemical research with respect to the prediction of molecular properties and biological activities, the quantitative modeling of reactivity has only been approached recently, and current models suggest that complex and specific parameterization is inevitable. As opposed to this, we report a simple machine learning model for predicting various reaction outcomes, such as yields and stereoselectivities. Being based on a solely structural input, our model should be transferable to diverse problems related to organic molecules.

Machine learning meets mechanistic modelling for accurate prediction of experimental activation energies

“…Accurate prediction of chemical reactions is an important goal both in academic and industrial research. [1][2][3] Recently, machine learning approaches have had tremendous success in quantitative prediction of reaction yields based on data from high-throughput experimentation 4,5 and enantioselectivities based on carefully selected universal training sets. 6 At the same time, traditional quantitative structure-reactivity relationship (QSRR) methods based on linear regression have seen a renaissance with interpretable, holistic models that can generalize across reaction types.…”

Machine Learning Enables Highly Accurate Predictions of Photophysical Properties of Organic Fluorescent Materials: Emission Wavelengths and Quantum Yields

“…To gain preliminary insights into the predictive powers of these ML models in conjunction with various molecular fingerprints, we first compared their mean absolute errors (MAE) for predicted absorption and emission wavelengths Morgan fingerprints show better performance, which implies that the representation of molecular structures by atom neighborhoods might be better for our purpose. In the recent study by Glorious et al 67 , the benefits of combining multiple fingerprints features (MFFs) as a composite input molecular descriptor was demonstrated. However, due to the extreme lengths of MFFs (more than 70,000 bits), the resultant increase of computation cost limits its application.…”

Molekulares maschinelles Lernen: Die Zukunft der Synthesechemie?

“…Erstere ist, wie die Entwicklung neuer Darstellungen (z. B. Mol2Vec, multimolekulare Fingerabdrücke, Graph‐basierte Modelle) immer wieder zeigt, für ein erfolgreiches ML unerlässlich und kann die Ursache für Misserfolg, Erfolg oder Fehlinterpretation eines Modells sein, während Letztere die endgültige Leistungsfähigkeit und die Lerngeschwindigkeit des Modells stark beeinflusst. Auch wenn dies adäquates Wissen aus vielen Bereichen erfordert, müssen wir damit beginnen, (selbst‐)kritisch über gewählte Methoden und fehlende Anpassung zu diskutieren, und sollten offen auf Mängel hinweisen, um die Verbreitung falscher oder schlechter Modelle zu vermeiden.…”