It’s common knowledge that experiments on mice or pigs, could hold the key to developments in the aging sphere. But, how about yeast? Google has the answer…
Researchers from Google and Calico Life Sciences are studying aging in yeast to develop a greater understanding of human gene expression.
Longevity.Technology: Understanding how we age and why we each age differently will enable those working in the Longevity sector to develop programmes, therapies and interventions that will promote independent and healthy living for as long as possible, as well as keeping age-related diseases at bay. Google, in partnership with Calico Life Sciences, has the big data experience to be able to really move forward with its machine learning model that successfully predicts gene expression in yeast. [1]
Calico is a research and development company with a mission to harness advanced technologies to increase understanding of the biology that controls lifespan, with the end goal of enabling people to live longer and healthier lives.
Back in 2103, with the launch of Google-backed Calico, a Time magazine cover story asked the question: “Can Google solve death?” [2] Clearly, the short answer is “no”, but the latest research certainly could take our understanding of how human genes work together as a complex system a further step forward.
The new study comes at a time when Google data and research could be used in a different way – to reshape future priorities in a post COVID-19 world, as discussed in Longevity.Technology’s latest interview with Longevity International CEO Tina Woods.
In the latest study, team at Calico Life Sciences created a genome-wide machine learning model to regulate gene expression in a species of yeast, which is a single-celled organism, by pioneering a way to make targeted changes to gene expression in yeast. This meant they could understand how aging works at a molecular level. In total, more than 200 experiments were carried out on a number of different strains of yeast; in each case, a single gene was “activated,” with expression levels of 6,000 genes measured eight times, leading to 20 million individual measurements.
Machine learning was used to model the entire data set as a system of differential equations which informed the prediction of how the other genes would react if one was “activated”.
Researchers then took this model and tested it against a validation set of 10 new yeast strains, finding that three out of the 10 predictions showed to be correct during experiments, including one gene that scientists had not previously identified.
“Our model was able to identify these without prior biological knowledge,” said Google Research’s Ted Baltz, “demonstrating that these techniques might scale to other domains or organisms that are much less well studied.”
While machine learning models have already been utilised in the understanding of inducible gene expression techniques, the new research goes further into the territory of understanding the “knock-on” effects further down the line of switching certain genes on or off. This could be key to understanding any unexpected negative effects of turning off or activating certain genes which are key to the aging process. For example, if you were to disrupt a certain genetic pathway in order to stop or reverse certain elements of the aging process, what would the impact be further down the line? How would the other genes in the human body then work together?
The hope now is that this greater understanding of how yeast ages can be applied to the aging process in much more complex organisms, such as humans. While at the early stages, the long-term aim is to develop a predictive framework to better understand how cells will behave over time – food for thought, the next time you’re making that pizza dough!
[1] https://ai.googleblog.com/2020/04/applying-machine-learning-toyeast.html
[2] http://content.time.com/time/magazine/article/0,9171,2152422,00.html
