Understanding metabolic processes through machine learning

December 27, 2018, Universität Düsseldorf
Understanding metabolic processes through machine learning
The researchers make use of various enzyme properties to predict kinetic parameters through machine learning. This allows the improvement of metabolic models and a better understanding of metabolism. Credit: David Heckmann

Bioinformatics researchers at Heinrich Heine University Düsseldorf (HHU) and the University of California at San Diego (UCSD) are using machine learning techniques to better understand enzyme kinetics and thus also complex metabolic processes. The team led by first author Dr. David Heckmann has described its results in the current issue of the journal Nature Communications.

The synthetic life sciences rely on a detailed and quantitative understanding of the complex systems in biological cells. Only if such systems are understood is their targeted manipulation possible. A system already well known is biological metabolism, in which many hundred enzymes are involved. However, a key aspect in this area, namely the individual activity of each , is only insufficiently understood in quantitative terms.

Together with his colleagues in California and Düsseldorf, Dr. David Heckmann, now in San Diego and former doctoral researcher under Professor Martin Lercher at HHU's Institute for Computational Cell Biology, has chosen a bioinformatics approach to get to the bottom of the properties of enzymes. For this, the researchers are using , a sub-field of artificial intelligence (AI), which is already used successfully in other areas, e.g. traffic management or automated translation. Machine learning algorithms have in recent years defeated their human opponents at chess, Go, and even poker.

Their approach allowed the researchers to identify important enzyme properties that are deciding factors for their activity. With these results, they can describe the kinetics of a large number of enzymes far more clearly than was possible with previous methods.

"Our model provides a fascinating insight into which enzyme properties influence their activity the most," says Professor Lercher.

"We can use this knowledge to model metabolic processes more precisely and to analyse the interaction of various components in cellular networks," adds Dr. Heckmann.

Explore further: AI heralds new frontiers for predicting enzyme activity

More information: David Heckmann et al. Machine learning applied to enzyme turnover numbers reveals protein structural correlates and improves metabolic models, Nature Communications (2018). DOI: 10.1038/s41467-018-07652-6

Related Stories

AI heralds new frontiers for predicting enzyme activity

November 16, 2018

Researchers from the Departments of Chemistry and Engineering Science at the University of Oxford have found a general way of predicting enzyme activity. Enzymes are the protein catalysts that perform most of the key functions ...

Scientists developed enzymes with remote control

November 26, 2018

Scientists have developed a method to enhance the activity of enzymes by using radio frequency radiation. The method requires a special complex of enzymes and magnetic nanoparticles. The particles can adsorb radio emission ...

Study lays foundations for future medicine design

September 12, 2018

A study by the Computational Biochemistry Research Group of the Universitat Jaume I (UJI) of Castellón, Spain, shows that enzyme activity depends on electrostatic properties, as opposed to mechanical ones. The scientists ...

Recommended for you


Please sign in to add a comment. Registration is free, and takes less than a minute. Read more

Click here to reset your password.
Sign in to get notified via email when new comments are made.