
SpaceX investor backs Toronto AI upstart's journey into the 'dark region' of genetic diseases
Brendan Frey set out, when he began the project to pick out a lead program for Deep Genomics, to prove that the artificial intelligence systems his lab has designed can identify new drug targets and find a winning candidate much faster than traditional methods. Now that they have zeroed in on an antisense oligonucleotide exon-skipping therapy for a subtype of Wilson disease — selected out of 2,400 ailments and 120,000 underlying genetic mutations — as their face case, the Toronto-based biotech is ready to delve into new frontiers with their AI tech.
Deep Genomics plans to spend roughly half of its new $40 million on clinical development of DG12P1, whose first-in-human trial is slated for 2021; and the other half on strengthening various components of its machine learning systems, from robotics to assays. Future Ventures, an investor in Tesla and SpaceX, led the Series B round while Amplitude Ventures, Magnetic Ventures, Khosla Ventures and True Ventures joined.
While they took a splice modulating approach for their first AI-discovered compound akin to what Biogen has with Spinraza, Deep Genomics is next focused on haploinsufficiencies — disorders where increased expression levels of one functional gene would be valuable. In other words, whereas the previous challenge lied in finding the right mutation to tinker, the new puzzle is how best to boost a known target.
It’s a relatively new application of oligonucleotides with no approved drugs, which is also being explored at Stoke Therapeutics.
“There are many different mechanisms that could be relevant, and so when you look at all these different mechanisms and all these different regions of the gene that you could design the oligo for, there are tens of thousands, or even hundreds of thousands of possible compounds,” says Frey, a star researcher that insilico’s Alex Zhavoronkov considers “without doubt in top 10 scientists in this field in the world.”
Frey elaborates:
That could be in the 5’ UTR, it could be in the intron, it could be in an exon, intronic sequences are very large, could be in a 3’ UTR, different kinds of mechanisms may be involved. It could be a matter of altering the upstream open reading frame, or it could be a matter of an intron or tension bottle neck, it could be a matter of changing the polyadenylation site, it could be a compound that alters secondary structures.
New tools will be required to test all these paths they can potentially travel down, and Deep Genomics’ team of 40-plus is already perfecting one model designed to predict polyadenylation patterns.
“Multilingualism is an important core value for us,” he previously said. “Everyone at the company has had AI training, and everyone at the company has done wet lab work. In fact I actually sequenced the genomic DNA for the Wilson target to validate it once we edited the cell line to put the patient mutation into the cells.”
In addition to deepening the clinical bench, Frey is also recruiting for the business development unit.
“Our pipeline is overflowing, and so we’re focused on signing some business deals,” he tells Endpoints News in the leadup to the annual JP Morgan confab in San Francisco.
That includes two additional programs for Wilson’s disease that, together with the lead program, would address up to a quarter of the 230,000 patients worldwide. There’s one other undisclosed metabolic candidate for which Frey plans to submit an IND this year, some compounds for retinopathy, and a dozen haploinsufficient projects in development, he says.