AI-driven Deep Genomics gets $180M to turn biology into informational medicines - MedCity News

Drug discovery and development is advancing by leaps and bounds as genetic medicines progress under a growing number of biotech companies. In the view of Brendan Frey, founder and CEO of startup Deep Genomics, while the drugs of the 20th century revolved around small molecule chemistry, drugs of the 21st century will be based on information: the sequence of letters in genetic code that determine health and disease.

Deep Genomics uses artificial intelligence to analyze genetic data. The technology of the Toronto-based company is feeding a pipeline of drug candidates, internal and partnered, that are advancing to clinical testing. Frey is now floating the ambitious goal of making Deep Genomics the world’s leader in producing and selling genetic medicines.

“Medicines are information now,” Frey said. “How do we convert the complex biology to a way to design new medicines? To dominate this entire area, we need a way to convert this complex biology to this informational medicine.”

On Wednesday, the company unveiled $180 million in financing to move it closer to its objective. The Series C round of funding was led by SoftBank Group, which invested from its Vision Fund 2.

Deep Genomics focuses on RNA, the molecules that carry instructions for making proteins. Continuing the comparison of biology to information, Frey likens the process of developing a Deep Genomics drug to writing a patch to fix a software bug. The company’s proprietary technology, called AI Workbench, applies AI and machine-learning techniques to discover targets for its medicines, then design therapies that can address those targets.

You could think of a Deep Genomics drug as a genetic software patch, Frey said. The drugs are oligonucleotides, which are short strands of nucleic acids. These genetic medicines are steric blockers, which is another way of saying that they bind to RNA and block that part from the cell’s protein-making machinery. Doing so alters the way in which a cell reads RNA, and in turn, corrects the mutation, Frey said. For haploinsufficiencies, which are diseases characterized by insufficient levels of an important protein, the patch can be programmed to get the RNA to ramp up protein production. If a mutation leads to a bad protein, the genetic patch can decrease protein production or ensure that the correct protein is produced. One example of a commercialized steric-blocking oligonucleotide is Spinraza, a Biogen drug approved for treating spinal muscular atrophy.

In 2019, Deep Genomics unveiled its first AI-discovered drug, an experimental treatment for Wilson disease. This rare, inherited disorder is caused by a genetic mutation that disables the liver’s ability to excrete excess copper. As a result, copper builds up in the liver and elsewhere in the body, leading to complications and in some cases, death.

The Deep Genomics pipeline has since expanded to nine additional programs. Frey aims to advance four into the clinic by 2023—all of them discovered and developed by AI Workbench. Some of the drugs are for rare diseases, but Frey said that the technology can also apply to more common diseases; the company’s programs include refractory gout and Parkinson’s disease. With the new financing, Frey said Deep Genomics aims to rapidly scale to 30 programs.

Deep Genomics is also seeking partners with larger biopharmaceutical companies. The startup currently has one partnership, an alliance with BioMarin Pharmaceutical that began last November. The four programs in that partnership are addressing undisclosed targets, but Frey said that the alliance includes disease targets that the San Rafael, California-based company brought to Deep Genomics as well as programs from the startup’s pipeline.

The reason for seeking partners is that AI Workbench is so productive that it generates more drug opportunities than Deep Genomics can pursue on its own, Frey said. Partners bring to the table capabilities that Deep Genomics does not have, such as expertise in a particular therapeutic area or experience running international clinical trials. AI aids in partnering decisions, as well. The analysis doesn’t remove all risks but it does inform where those risks are. For example, some programs may require a longer clinical trial to prove efficacy while others can show their value in earlier stage studies. Frey said analyzing those risks helps Deep Genomics make rational decisions about who is best suited to develop a particular program.

Deep Genomics has plenty of competition in the genetic medicines space, but Frey points to a feature that sets his company’s drugs apart from gene therapies and gene-editing medicines. Unlike those therapies, which make permanent changes, a Deep Genomics genetic patch can be removed. That way, if serious problems emerge, treatment can be stopped to reverse the adverse effects.

Another difference is the value that’s created by the Deep Genomics platform. Patents on CRISPR are now 10 years old, and the technology will eventually become a commodity, reducing its value, Frey predicts. Because the Deep Genomics technology learns and improves each time it is used, he said its value will increase over time. However, CRISPR editing drugs and gene therapies aren’t Deep Genomics’ only competition. Other startups with RNA platforms that offer features to those of the Deep Genomics technology include Strand Therapeutics and CAMP4 Therapeutics.

Prior to Wednesday’s funding announcement, Deep Genomics’ most recent financing was in 2020, a $40 million Series B round led by Future Ventures. The company’s latest financing included earlier investors Fidelity Management & Research Company, Canadian Pension Plan Investment Board, True Ventures, Amplitude Ventures, Khosla Ventures, and Magnetic Ventures. Frey said Deep Genomics was looking for investors with a long investment horizon. An IPO could come “when the time is right,” he added. That timing is partly dependent on the progress of the Deep Genomics pipeline. Frey said that the company won’t go public until it has four programs in the clinic that demonstrate what AI Workbench can do.