Some of the best scientists are actually supercomputers

Everything is super! Supercomputers will soon be able to solve mysteries related to complex cellular processes using machine learning. Scientists from both universities and the industry are hard at work developing and training computers to become superfast super scientists.

She is only around one year of age, but she has already identified around 400 potential molecules that can be produced in genetically optimized yeast cells.

“With Lila, we can work in a systematic way to get to the protein or the molecule. For instance, when selecting an enzyme, Lila can go through the literature and use machine learning approaches to study publications and accessible information on a specific protein. It’s not equivalent to a scientist reading a paper, but it is scalable and can use sophisticated statistical methods says Darren Platt, Vice President of Data Science at Amyris.

Lila is Amyris new supercomputer, who is still growing up. She has a lot to learn as scientists feed her with input and boost her “intelligence” with machine learning. In this way, she can figure out complex routes and correlations within the cell:

“Lila can now design the pathway, because she knows which molecules are made by the same enzymes. We then test a lot of samples, and Lila looks at how well they are doing and do classic machine learning by recombining the factors to get to the optimal design.” says Amoolya Singh, Director of Scientific Computing, who leads the team building Lila.

But the scientists will take Lila to the next level:

The next steps of strain optimization are much more complicated, this is where Lila will deal with optimizing how the carbon flows through the cell and other metabolic engineering issues.

At the Novo Nordisk Foundation Center for Biosustainability, scientists are also working on modelling all the routes of the cell to combine them to big road maps. The goal is to get intelligent models like Lila that can tell you what will work. This compares to having a roadmap of a city with heavy traffic in some parts. You know, you want to go from A to B, but have trouble finding the fastest and most fuel efficient way to go there. But here the model will give you different suggestions just like Google Maps.

Biology in the cloud

Just recently, Senior Researcher Nikolaus Sonnenschein, received a two million grant from the Villum Foundation to the experimental project ‘Hands-off biology: towards full automation of life science experiments in the cloud’ that explores whether life science can be conducted entirely from a computer without ever touching a pipette or entering a laboratory.

"“We are working towards full automation of life science experiments in the cloud. This will enable scientists to spend more time on data interpretation and less on dull repetitive tasks such as pipetting,”"
Nikolaus Sonnenschein, Senior Researcher at DTU Biosustain

“We are working towards full automation of life science experiments in the cloud. This will enable scientists to spend more time on data interpretation and less on dull repetitive tasks such as pipetting, says Sonnenschein.

A growing concern among scientists is the problem of irreproducibility. In 2012, an article in Nature Communications stated that pharma and biotech companies can only reproduce between 11 and 25 percent of the high-impact research papers in the field of cancer research.

“Manual experimental operations and data reporting have significant sources of error. Data can provide a permanent record of every experiment and tell researchers how to improve their technique and help to ensure that their execution of protocols is consistent,” according to Sonnenschein.

Change your mindset

In order for supercomputers like Lila and projects such as ‘Hands-off biology: towards full automation of life science experiments in the cloud’ to be successful, it requires a change in mind among scientists.

The time where students were only taught classical skills such as how to classify plants and animals in biology classes is long gone.  Data analysis and data mining will become an essential part of the curriculum within this field in the future. However, this development should not make people neglect biological knowledge.

Computers and robots have a huge potential to teach us a lot about biology, but they need to be provided with the right details. Even tiny details matters and you only get good answers if you are capable of asking the right questions, emphasizes Sonnenschein.

While, the white lab coat can be spared from time to time, one thing will remain just as important as always. The ability to be critical.

Lila and her siblings can give very accurate results when boosted with intelligence, but they are only really useful for the understanding of science and biology, if scientists are able to answer the million-dollar question:

“How did you come up with the result?”