Deep Genomics

7,000 rare genetic diseases. Only 5% with approved treatments. 95% of patients—without hope.

Professor Brendan Frey stared at these numbers, over two decades since the Human Genome Project revealed 20,000 human genes and millions of disease-causing genetic variants. Yet the pharmaceutical industry couldn’t translate this knowledge into proportional therapeutic advances.

Traditional drug discovery: Human intuition. Sequential trial-and-error. 4-6 years identifying a single drug candidate. 10-20% success rates. Billions wasted while patients—many children—deteriorated.

Frey, University of Toronto Professor of Engineering and Medicine and deep learning pioneer, spent a decade building increasingly sophisticated AI systems. By 2014: superhuman capability. His AI interpreted genetic mutations more accurately than human experts.

Industry response: Silence.

When pharmaceutical companies didn’t immediately adopt his breakthrough, Frey made an audacious decision: to commercialize it himself. Transform AI predictions directly into life-saving medicines. In 2015, he founded Deep Genomics.

This represents why Deep Genomics, a Toronto-based AI-driven genetic medicine company, has won 2026 Global Recognition Award for developing the AI Workbench that nominated DG12P1 as industry’s first AI-discovered oligonucleotide therapeutic for Wilson disease in just 18 months (versus traditional 4-6 years), achieved 70% preclinical success rates (versus pharma’s 10-20%), developed BigRNA foundation model enabling tissue-specific oligonucleotide design, and established partnerships with BioMarin Pharmaceutical on four rare disease programs.

Technical Innovation

Traditional oligonucleotide design: Trial-and-error. Test compounds sequentially. Takes years.

Deep Genomics‘ AI Workbench changes everything. The platform integrates deep learning models trained on RNA biology with massive biological datasets. It predicts how genetic mutations cause disease at the molecular level. Designs oligonucleotide therapies restoring normal cellular function by modulating RNA processing.

Target identification algorithms scan thousands of diseases and hundreds of thousands of pathogenic mutations to identify precise disease-causing mechanisms. Oligonucleotide design tools explore tens of billions of potential sequences. Off-target prediction models assess genome-wide effects, eliminating toxic compounds before expensive synthesis. Iterative learning loops feed every experimental result back into the model, continuously improving predictions.

DG12P1 exemplifies platform power. Wilson’s disease: a devastating genetic disorder causing toxic copper accumulation in the liver and the central nervous system. The AI Workbench scanned over 2,400 diseases and 100,000+ pathogenic mutations, identified the Met645Arg mutation that causes loss of the ATP7B copper-binding protein, predicted a precise molecular mechanism (aberrant RNA splicing), designed oligonucleotides to correct this defect, screened thousands of compounds, and identified DG12P1 as the optimal candidate.

Timeline: 18 months from target identification to drug candidate versus traditional 4-6 years. Success rate: 70% versus pharma’s typical 10-20%. The September 2019 DG12P1 nomination demonstrated that AI could fundamentally transform drug discovery from hypothesis-driven trial-and-error to prediction-driven systematic exploration.

BigRNA (September 2023) represents the latest evolution. Transformer neural network foundation model trained to predict biological mechanisms

regulating RNA expression tissue-by-tissue. Result: tissue-specific oligonucleotides acting selectively in target tissues, avoiding off-target effects. Validation studies of 14 genes, including Wilson disease and spinal muscular atrophy, showed that BigRNA consistently designed highly effective compounds. The company evaluated over 9 billion molecules against 1 million targets, generating terabytes of proprietary biological data, creating competitive moats.

Leadership and Market Strategy

Brendan Frey brings an exceptional pedigree: University of Toronto Professor, Vector Institute co-founder (Canada’s national AI research institute with $150+ million funding), Royal Society of Canada Fellow, AI pioneer with fundamental contributions to deep learning. His 2004-2014 journey building AI systems, culminating in superhuman mutation interpretation, followed by his decision to commercialize when the industry didn’t adopt, demonstrates extraordinary persistence and entrepreneurial courage.

Exceptional advisors include Yann LeCun (Turing Award winner, Meta Chief AI Scientist), Richard Scheller (former Genentech EVP Research), Jennifer Cook (former Grail CEO), and Steve Jurvetson (Future Ventures founder, Tesla/SpaceX board member), who led $40 million Series B in January 2020.

August 2021 brought $180 million Series C led by SoftBank Vision Fund 2—one of the largest AI drug discovery financings. Total disclosed funding: approximately $230+ million.

In November 2020, BioMarin’s collaboration validated the platform on a commercial scale. This leading rare disease pharma (market cap $15+ billion, multiple approved drugs) selected Deep Genomics’ AI Workbench to identify oligonucleotide drug candidates in four rare disease indications. The partnership model enables scalability, as the platform supports unlimited partners simultaneously, generating discovery collaboration revenues, milestone payments, and potential royalties while continuously building proprietary datasets that continuously improve capabilities.

Industry Impact

Deep Genomics addresses the healthcare crisis in which, despite the Human Genome Project’s completion, revealing the genetic causes of thousands of diseases, the pharmaceutical industry has failed to translate this knowledge into treatments. Over 7,000 rare genetic diseases exist, with only ~5% having approved treatments, leaving 95% of patients without hope because traditional drug discovery cannot systematically identify which millions of genetic variants are targetable and which therapeutic interventions effectively correct defects.

By enabling AI-powered systematic exploration evaluating billions of potential oligonucleotide sequences against millions of genetic targets computationally before expensive synthesis, Deep Genomics reduces drug candidate identification from 4-6 years to 12-18 months while achieving 70% success rates versus pharma’s 10-20%—fundamentally transforming genetic medicine economics from opportunistic development targeting largest patient populations to systematic coverage addressing even ultra-rare conditions.

The impact: bringing treatments to 95% of rare disease patients currently lacking options and establishing computational genetic medicine as a standard approach for monogenic and polygenic disorders affecting millions worldwide. This is why Deep Genomics has earned recognition for innovation that doesn’t just incrementally improve drug discovery—it fundamentally reimagines how humanity translates genetic knowledge into life-saving medicines.