Machine learning aids medication developers and brewers by predicting metabolism.
Scientists at the Francis Crick Institute have created machine learning algorithms that can predict yeast metabolism based on its protein concentration. Brewers may be able to have more control over the flavour of their beer as a result of the results, and scientists may be able to personalise therapies for metabolic problem patients in the future.
Metabolism is a set of chemical events through which organisms turn foods into energy and necessary chemicals. When yeast metabolises sugar in the absence of oxygen, it 'ferments,' producing alcohol, acids, and gases, as well as flavour compounds, which are what give bread, wine, and beer their distinctive flavours.
Metabolism produces hundreds of tiny compounds called metabolites within a cell. Although yeast and humans are evolutionarily different, many of these metabolites are same and produced in comparable ways. However, the processes that control metabolism have remained a mystery until now.
Machine learning algorithms can predict the metabolism of brewer's yeast (S. cerevisiae) to a great extent if they are given huge volumes of protein expression information, according to a new study published in Cell Systems.
"We now have a better understanding of what controls metabolism thanks to machine learning, which is good news for brewers looking to make the perfect pint, or for biotechnologists using yeast to produce vaccines and other medically important proteins," says Aleksej Zelezniak, first author of the paper and researcher at the Crick, who recently moved to Sweden to start his own research group at Chalmers University of Technology.
Protein-metabolite interactions
Until date, scientists have been split on whether metabolism is self-regulating or controlled by changes in gene expression, partially due to the failure of existing methods to find a substantial association between the read-out of genes (proteins) and metabolites.
Researchers examined enzyme expression in 97 different strains of S. cerevisiae, which are known to have diverse metabolisms, and linked it to variations in metabolite concentrations.
They created machine learning algorithms that could detect intricate correlations between gene expression changes and the metabolites generated. They discovered that metabolism is governed by a complex network of enzymes working together, with no single enzyme having a significant impact on its own.
"The link between enzyme expression and metabolism in yeast is so intricate that prior models have failed to detect it," explains Markus Ralser, the paper's senior author and group leader at the Crick. "Changes in cellular metabolism are closely linked to conditions that worsen with age, such as diabetes, cancer, and neurodegenerative diseases. The ability to predict metabolism in basic cells like yeast cells is a significant step forward in the effort to predict metabolism in human tissues.
"Tech titans like Amazon and Facebook routinely utilise similar computational technologies. Instead of utilising them to personalise adverts or recommend pals, we've used them to anticipate the metabolism of yeast cells. These findings help us understand not only the basis of beer flavouring, but also several human metabolic problems."
Beer to individualised medicine
The researchers hope to apply their discoveries in yeast cells to patients with metabolic illnesses in the next few years.
"It may seem unusual to non-biologists that we can translate our understanding of yeast to people," Aleksej explains, "yet many key aspects of what we know about human biology emerged from yeast research."
"We're currently working on expanding our algorithms to include information on a person's metabolism based on the proteins found in their blood. This data could aid doctors in determining which therapy choice is optimal for each patient."