About tough problems and smart solutions: Let AI do the lab work

Author:

Dr. Valeria Scagliotti 
Sustainability Consultant for Life Sciences at SustainLABility

Key topics: Sustainability, Environmental Footprint, Climate Change, IPCC Report, Greener Future

Category: Sustainability

Date: 10/10/2022

AI is your new colleague in the lab.

It is when human intelligence is facing an unresolvable problem that we need to step aside and let computers do the work. We all know that computers work more efficiently than humans while also ruling out the risk of human failure that just lies in our nature. For the longest time computers have been doing what they were programmed to do. They just act faster and more precise than people are able to.

But not long ago a new foe – or friend, if you will – appeared. Many fear it, and its rise against humanity as it becomes more powerful, while others greatly welcome it as the problem solver we all desperately need in our sometimes too complicated world. This newcomer amongst revolutionary tech? Artificial Intelligence (AI). 

AI is not meant to work only with set information that is given to it, but to also actually understand and learn, and thus constantly evolve and become a better version of itself. AI can find and understand patterns among information and data it is fed, and it can learn from experience like success, or failure. AI basically works like a human brain but at a significantly higher performance rate.

 

The Protein Problem

Proteins are the base of everything in life – the good, and the bad. Proteins are basically the code writing life itself. For years, molecular biologists have been trying to decode protein structures to find the protein codes that are responsible for certain diseases. You may ask yourself why. Well, the answer is clear: if you can map out a problem in great detail, you are more likely able to solve it, right?

How come it is so difficult to analyse those structures? Put simply, proteins not only have huge, 3-dimensional structures, they are also folded. There are so many possibilities that you could end up trying to calculate a structure for years, or even decades, and would still not find the answer.

In 1972 the American biochemist Christian Anfinsen was awarded the Nobel prize for discovering that amino acid sequences determined the 3D-structure of proteins. From that moment on, scientists all over the world devoted their time and energy to solve a problem that just seemed insoluble.

So far around 170,000 protein structures have been experimentally determined by measuring them – a process that takes far too much time. So, the idea of calculating them instead, came to mind.

 

AI is on the rise and it’s here to save our lives

For the past 50 years scientists have been devoting themselves to the protein problem, one of them being Demis Hassabis. Since his student days in Cambridge in the 1990s, the computer scientist has been fascinated by this problem and finding a way to solve it.

And low and behold, he succeeded! In 2010 he founded the company DeepMind that is now owned by Alphabet Inc., the parent company of Google.
DeepMind research is devoted to machine learning, programming AI that can play video games, or even solve life-changing problems. The program solving the protein folding problem is called AlphaFold.

DeepMind recently gained a lot of attention when taking part in the infamous challenge ‘Critical Assessment of Structure Prediction’, also known as CASP. The results of the contest were released at a press conference on 30 November 2020. AlphaFold outperformed the other teams by such a high rate that there is probably no scientist left out there that can say they have never heard of the hot, new game changer in the lab.

 

AI is your new colleague in the lab

AlphaFold was given the data of the 170,000 protein structures that are known so far and uses this information to calculate new structures. With AlphaFold being able to solve a problem within minutes - while scientists need years of research without even finding the solution in the end - AI might become your new favourite lab colleague in the future. And it might save us all in the end … or destroy humanity, who knows! Anyway, we are excited to find out what the future of artificial intelligence will look like and we cannot wait to see how it will change science as we know it.

 

Your daily dose of AI

Did you know that AI has been around for a long time in our everyday lives? Not only does it make your life easier with the programs or apps you use on a daily basis, you are also participating in its development. Yes, you have been training AI the whole time without knowing it.

Did you ever solve a ‘captcha’? You know, those barely readable words, or weird combinations of letters and numbers. Well then, you might have been training AI that can identify text to scan old books existing only in print format to make the text digitally available. Pretty cool, right? Or have you ever solved a captcha asking you to click on certain items like vehicles, traffic signs, or buildings on an image consisting of blocks? Then you might have been training AI for autonomic driving. Isn’t this exciting? You are part of the development of AI all along!