Deep Genomics, based in Toronto and focused on drug discovery through artificial intelligence. Their platform combines automation, biomedical knowledge and data with artificial intelligence for maximum effect.
Wave Life Sciences and Deep Genomics have come together to identify new approaches for treating genetic neuromuscular disorders. Both organizations will utilize Deep Genomics’ chemistry platform to validate targets and clarify potential interventions across phenotypes.
Machine Learning
Research indicates it takes the pharmaceutical industry approximately $2.6 billion and 10 years on average to develop a new drug, much of which comes from trial-and-error procedures required to identify promising therapies. Machine learning holds great promise to shortening this R&D process by narrowing the number of candidates for investigation, predicting biological outcomes of mutations, and recognizing clinically-relevant variants from genomic data.
Deep Genomics’ machine learning technology can detect disease-causing mutations by analyzing their effect on gene expression and splicing regulation, then using this knowledge to develop oligonucleotides to either increase or decrease gene expression or alter splicing regulation. Its design works on numerous biological pathways while taking advantage of information sourced from multiple sources like its own genetic and genomic data sets, public genomic databases and clinical archives as well as published literature.
Wave and Deep Genomics’ technology has been successfully employed to identify a pipeline of RNA therapeutics, including splice correction therapies for genetic neuromuscular diseases like Wilson disease and spinal muscular atrophy as well as metabolic diseases like gout, refractory gout, frontotemporal dementia and Niemann Pick disease type C. Through collaboration, their machine learning platform will identify cause-and-effect relationships specific to neuromuscular targets involving splicing regulation.
Recent developments have led to increased interest in machine learning for medicine. These include large-scale approaches to integrate data from diverse modalities – imaging and genetics – using deep learning models; multiple kernel learning (MKL), which combines different modalities using different kernel functions; and subtler uses of ML as part of PRS systems or linear predictors.
Geneticists have demonstrated that machine learning (ML) improves disease prediction with respect to traditional, linear models using SNPs that correlate with causal variants. Yet its predictive power for complex trait genetics still falls below 1, despite increasing genome-wide genotype and phenotype information being made available, likely due to limitations posed by traditional linear predictors that do not capture nonlinear effects and interactions that contribute to complex diseases.
Oligonucleotide Discovery
Oligonucleotide drug development involves targeting genetic determinants of disease in DNA or RNA. To facilitate this process, AI Workbench utilizes deep learning, automation and advanced biomedical knowledge to accurately identify targetable mechanisms and potential therapeutic candidates. Furthermore, this process generates on-target effect data such as cell viability or animal toxicity that may serve as biomarkers in clinical trials.
Deep Genomics has signed several partnerships to put its AI platform to the test, such as one with Wave Life Sciences that will focus on finding novel therapies for genetic neuromuscular disorders. Wave will use its AI Workbench to prioritize targets within genes underlying these conditions while its chemistry platform validates and clarifies these potential interventions across different phenotypes.
Wave is also applying its expertise in designing stereopure oligonucleotides specifically tailored for gene therapy applications to create molecules designed to modulate protein expression levels and thus alleviate disease symptoms – this approach mirrors what pharmaceutical companies take when developing small molecule drugs.
Most conventional medicines work by altering the behavior of proteins – blocking receptors, inhibiting enzymes or altering downstream signaling pathways. But many inherited diseases involve misfolded or mutated proteins which cannot function correctly, leading to muscle weakness symptoms as a result.
Deep Genomics platform’s ability to detect such rare mutations could pave the way for developing novel treatments for hereditary diseases and cell reprogramming, an approach already used successfully by several hepatitis C patients.
Deep Genomics can become one of Canada’s next great biotech success stories with the right strategy and execution, however to do so they must invest in machine learning research programs and academic programs to maintain thought leadership position as well as core AI platform that will ensure both short-term partnerships with pharmaceutical firms as well as long-term growth success.
Preclinical Development
Deep Genomics’ AI Workbench platform quickly identifies potential oligonucleotide therapies to target genetic determinants of disease at either the level of RNA or DNA. Furthermore, this system generates on-target and genome-wide off-target effect data for every compound identified, along with cell viability and animal toxicity information for every molecule discovered – data which Deep Genomics states is then integrated into an iterative machine learning model to increase predictive accuracy.
Deep Genomics currently has several preclinical candidates under development, such as DG12P1, an antisense oligonucleotide designed to treat autosomal recessive Duchenne muscular dystrophy. As Wave’s first antisense oligonucleotide that targets one specific RNA target directly and with the aim of correcting mutations and restoring normal gene function while potentially halting disease progression, its development represents a milestone.
Wave’s second clinical portfolio candidate is WVE-006, an antisense nucleic acid therapeutic targeting an individual RNA target within the cell machinery to produce proteins, commonly referred to as RNA editing. A Phase 1b trial has already demonstrated proof of concept with WVE-006 helping correct severe disease symptoms; further investigations will take place to validate efficacy and safety of this splice correction therapy.
Another project underway is DG12P2, an antisense oligonucleotide targeting exon 51 of DMD to stop production of an altered exon, believed to cause muscle weakness and motor control deficits. The company plans on beginning Phase 1b trials this year.
Armed with their sturdy platform, the team is now focused on building partnerships with pharmaceutical companies to advance their drug programs. Furthermore, AI Workbench was developed as a way for biopharmaceutical firms to convert RNA sequences into potential therapeutic compounds for further analysis and use by biopharmaceutical firms.
As it transitions into commercialization, a biotechnology company must remain mindful of data biases while maintaining strong machine learning research and thought leadership. Ultimately, this will be crucial to its short-term ability to secure pharmaceutical deals and long-term success as an AI-powered biotechnology firm; accessing quality genomic data as well as diverse research populations are vitally important in order to maximize machine learning’s predictive power.
Commercialization
Wave’s technology aims to make drugmakers’ jobs simpler by helping them discover more effective therapies for genetic diseases. Their machines use artificial intelligence technology to detect patterns that humans cannot see and incorporate their insights into drug production processes. Frey anticipates that in time almost all drugs will be produced using AI as this saves both time and money by offering more effective therapeutic approaches for genetic ailments.
Deep Genomics of Canada recently joined forces with Wave Life Sciences of Cambridge, MA, to test their machine learning technology’s efficacy in drug development. Together they signed an agreement that will identify novel therapies developed by Wave for treating genetic neuromuscular disorders through targeted splicing regulation regulation targets and validation using Wave’s chemistry platform as part of an effort to expand their pipeline of rationally designed oligonucleotides.
Deep Genomics announced an important milestone with this deal; although its exact value wasn’t revealed. Established in 2015, Deep Genomics has raised over $100 million and published numerous scientific papers using machine learning-based approaches to finding disease triggers.
Deep Genomics faces fierce competition from startups and multi-national pharmaceutical companies that seek to establish internal machine learning capabilities of their own. As such, it is imperative that Deep Genomics invests in its core machine learning technology while at the same time being able to deploy this in its drug development process effectively.
Wave has the opportunity to become a market leader in genomics research and forge successful partnerships with major pharmaceutical companies by building upon its strong foundation. To do so, its leadership must consider carefully how resources are allocated between its machine learning research/academic arm and drug development and partnership efforts; failure could compromise industry leadership as well as opportunities to strike lucrative deals in this competitive space.
