PREFER: A New Predictive Modeling Framework for Molecular Discovery

Machine-learning and deep-learning models have been extensivelyused in cheminformatics to predict molecular properties, to reducethe need for direct measurements, and to accelerate compound prioritization.However, different setups and frameworks and the large number of molecularrepresentations make it difficult to properly evaluate, reproduce,and compare them. Here we present a new PREdictive modeling FramEwoRkfor molecular discovery (PREFER), written in Python (version 3.7.7)and based on AutoSklearn (version 0.14.7), that allows comparisonbetween different molecular representations and common machine-learningmodels. We provide an overview of the design of our framework andshow exemplary use cases and results of several representation-modelcombinations on diverse data sets, both public and in-house. Finally,we discuss the use of PREFER on small data sets. The code of the frameworkis freely available on GitHub.

Automated summary

Machine-learning and deep-learning models are widely used in cheminformatics to predict molecular properties and accelerate compound prioritization. However, evaluating and comparing these models is challenging due to different setups, frameworks, and molecular representations. In this study, we present a new framework called PREFER, based on Python and AutoSklearn, that allows comparison between different representations and machine-learning models. We provide an overview, examples, and discuss the use of PREFER on small data sets. The framework's code is freely available on GitHub.