In 1995, our founder Brendan Frey helped to lay the foundation for the highly successful artificial intelligence technology called “deep learning”. That’s the technology that’s changing everything, from farmers growing better crops to cell phones converting speech to text. In 2002, he faced a family medical crisis that led him to recognize that good drugs are based on good biology, but biology is too complex for humans to understand and accurately predict, especially at scale. So, Dr. Frey decided to focus on that problem. For the next 13 years, he and his team built artificial intelligence that could be used to successfully discover the genetic causes of disease and new therapeutic approaches to treat them. Their discoveries made the front pages of journals like Nature Magazine, and international newspapers, such as the New York Times.
By 2015, the artificial intelligence technology was so powerful that it was time to bring it to patients. The team founded Deep Genomics, whose mission is to serve patients by building and using artificial intelligence to discover and develop better treatments for genetic diseases, both rare and with large prevalence. In 2018 this artificial intelligence technology discovered how a specific Wilson Disease-causing mutation works, something human scientists hadn’t yet figured out (you can read our paper on that discovery here).
We have prepared a short video summarizing how Deep Genomics uses artificial intelligence to discover potential therapies that target these genetic diseases.
Genetics
To understand how we use artificial intelligence to design potential therapies, let’s first review some basic concepts in genetics.
Our genes are made of DNA which is present in nearly every cell of our bodies. Each gene is usually present in two copies, one from each parent. The DNA provides the instructions to make different proteins which are responsible for the various functions in the body. For example, proteins are important for hair and eye color, growth, brain development, movement and metabolism - the conversion of food into energy or body fat and muscle.
The process of building proteins begins with DNA. Cells first read the instructions contained in DNA, which are written using the genetic alphabet of A, C, G and T, to make a copy called RNA, which is written using a similar genetic alphabet. This process is called transcription. Then, cells read the instructions contained in the RNA to make proteins, using a process called translation. You can think of RNA as the molecule that carries the message between the DNA and the protein.
Many rare diseases are caused by changes to the letters in the DNA, called genetic mutations. Mutations in the DNA get transferred to the RNA and this can lead to a protein that doesn’t work properly, or can even prevent a protein from being made at all. Many rare genetic diseases are caused by these types of mutations.
Deep Genomics uses artificial intelligence to discover potential therapies that target these genetic diseases.
How artificial intelligence helps us discover therapies
For a single disease, there may be thousands of disease-causing mutations to look at, and hundreds of different ways of trying to fix the problem. On top of that, there may be hundreds of thousands to millions of potential drugs to search through, but only a few that work.
Sifting through so many possibilities to find those that may help is incredibly time consuming, and complicated, for human scientists, and that’s where our artificial intelligence technology offers advantages.
There are two main ways we use our artificial intelligence to help us develop potential medicines.
1. Using artificial intelligence to discover targets
First, we use artificial intelligence to rapidly examine all the possibilities and identify disease-causing mutations and ways of fixing the genetic problem. This is called target discovery and because there may be thousands of mutations and hundreds of different ways to fix the problem, having a computer do the work is a big boost.
2. Using artificial intelligence to design therapies
Second, we use artificial intelligence to design therapeutic candidates. For a specific disease, our artificial intelligence can assess hundreds of thousands to millions of different potential targeted therapies to find the ones that are most likely the best. These are then verified in our lab. Again, a big boost.
The class of therapies that Deep Genomics is currently developing are called steric-blocking oligonucleotides.
Steric blocking oligonucleotides
Steric blocking oligonucleotides are short stretches of special DNA or RNA that attach to a specific place in the RNA. By doing so, they modify that natural way in which cells process the RNA and build proteins. They don’t make any permanent changes to the DNA.
A steric blocking oligonucleotide can potentially restore production or the amount of the protein.
