Home / Blog / Beyond CRISPR, leveraging on Artificial Intelligence

Beyond CRISPR, leveraging on Artificial Intelligence

The sequencing of the genome looks like a set of program instructions. CRISPR provides a good way to change some of the code but what are the consequences? Image credit: MIT Technology Review

The code of life, the DNA, can be sequenced in a matter of days (that is our human code, there are “shorter codes” like the ones of some viruses but also much longer ones, like the one of a plant, the Paris Japonica, that is 50 times longer than our -hence do not believe we are the most complex beings on planet Earth!). It used to take months (the first sequencing took 7 years) and it is getting shorter, and cheaper as time goes by.

CRISPR/Cas9, a technology/process borrowed from bacteria, is now allowing researchers to edit the DNA, replacing a sequence of codons with a different one, in practice doing what a programmer does to fix a problem in a software program or to change/enhance its features.

There are several problems of course in spite of the fact that this technology has proven extremely easy and effective (or possibly because of this).  The fact is that the code of life is much more complex than a software program, at least a classic one (a different story applies to self constructed software as the result of deep learning and convoluted networks where we also start to experience “issues”).

There is, normally, no direct one to one correspondence between a set of instructions -codons- and the manifestation of those instruction. It is not like: you have some instructions set determining that your hands have five fingers. Let’s change those instruction and create a hand with 6, or 4, fingers. Yes, in principle you would be able to generate a hand with 4 fingers but at the same time you would create a sterile individuals. Those genes that result in a hand with 5 fingers also ensure that the individual is fertile. That is why you won’t find people with hands having a different number of fingers. Those that evolution generated where not able to reproduce, hence that evolution never took roots.

Another problem is that the CRISPS/Cas9 is like an edit command telling: “search all occurrences of CGAT – CGTA – GCTC and replace them with TCGA – TCGA – CCTA”. This may lead to unexpected results, since what you want to do is replace only one (or few) occurrence of that codons sequence, not all of them.

So the question facing researchers using CRISPR/Cas9 is: “what is going to happen when I make the change? Sure I will get the expected change but what else, unexpected, is going to happen?”

Scientists are looking at ways to answer this tricky question and are experimenting with Artificial Intelligence. Microsoft has built a machine learning tool to help in CRISPR/Cas9 design.

Using a deep learning approach the tool can discover what is the correlation between the editing applied to the genome and the consequences at a global scale, basically linking the genotype with the phenotype. In the future this may provide a crucial tool to fix problems appearing at the phenotype level by tweaking with the genome (it will also show that some problems do not have a solution with this approach, since fixing something -like changing the number of fingers in a hand- will definitely ruin something else).

It is somewhat peculiar that to answer a question we do not know how to tackle we are using a tool we don’t know how it is working! Indeed, we know by experience that it is working but we actually do not know “why” it is working. In a way it is similar to the issues we face in quantum physics, where we apply equations that happens to return results in accordance with our experiments although we do not understand why this is so!

On the other hand, it is also true that we have been solving Nature puzzles by using our brain and as a matter of fact we do not know, yet, why it works….

About Roberto Saracco

Roberto Saracco fell in love with technology and its implications long time ago. His background is in math and computer science. He's currently the Chair of the Symbiotic Autonomous Systems Initiative of IEEE-FDC. Until April 2017 he led the EIT Digital Italian Node and up to September 2018 he was the Head of the EIT Digital Industrial Doctoral School. Previously, up to December 2011, he was the Director of the Telecom Italia Future Centre in Venice, looking at the interplay of technology evolution, economics and society. At the turn of the century he led a World Bank-Infodev project to stimulate entrepreneurship in Latin America. He is a senior member of IEEE where he leads the Industry Advisory Board within the Future Directions Committee. He teaches a Master course on Technology Forecasting and Market impact at the University of Trento. He has published over 100 papers in journals and magazines and 14 books. He writes a daily blog,  http://sites.ieee.org/futuredirections/category/blog/, with commentary on innovation in various technology and market areas.

Leave a Reply