DeepTech in Drug Discovery
“The drug discovery process consists of many phases and often takes decades. The probability of failure is very high at every stage. In preclinical phases the failure rates are over 99%. Our AI can be used in every one of these phases and in some cases lead to superhuman results. Our AI is exceptionally good at finding the molecular targets in specific diseases and inventing new chemistry.”
Alex Zhavoronkov, CEO, Insilico Medicine
DeepTech companies combine cutting edge science and advanced engineering with the objective of making a profound impact on humanity. One of the sectors most impacted by DeepTech is AI for Drug Discovery. This is the second article in our DeepTech Series. In this article we will focus on AI for Drug Discovery and profile four top tier companies in this sector. Drugs are among the most effective ways to combat disease. Drugs are used to fight cancer, relieve pain, prevent cardiovascular disease, alleviate mental illness, cure infectious disease, and much more. In contrast to other areas of human endeavor, drug discovery has not become faster and cheaper with time. Using AI for drug discovery could help change this.
Any drug, when introduced into the human body, interacts with unintended targets. This leads to side effects that require a huge amount of costly trial and error and often doom entire drug development programs. Predicting how a molecule will interact with all possible protein targets and designing molecules to interact with desirable proteins while avoiding toxic proteins can increase efficiency and mitigate downstream clinical failures. This is a non-trivial problem that requires a combination of computational biophysics and AI methods.
In 2014, Cyclica’s scientists started working on a solution to this problem. They set out to develop a computer algorithm that could test any given molecule against all known proteins in the human body and identify those that would be affected. In 2017, Cyclica turned this technology into the Ligand Express™ platform. This platform helps medicinal chemists to quickly identify mechanisms of drug action to improve the prospect of newly discovered drugs. It also helps reduce the amount of laboratory work and the number of animal experiments needed to bring new drugs to market.
“At our core, we are a biophysically-oriented company that leverages AI to augment our biophysical simulations to conduct small-molecule proteome screening for drug discovery, all at the fingertips of the end user who are medicinal chemists, biochemists, and screening scientists.”
Naheed Kurji, CEO, Cyclica
Scientists at Cyclica believe that a deep understanding of the chemistry and biophysics of drug/protein interactions is fundamental to successful drug discovery, and that approaches based purely on data and AI can only take scientists so far. In 2017, Cyclica set out to augment its biophysical approach with deep learning methodology. That led to the development of MatchMaker™which combines biophysical simulation with deep learning to achieve a quantum leap in accuracy and throughput for Ligand Express.
With the MatchMaker powering Ligand Express, large parts of the drug discovery process can now be performed in silico, greatly reducing both time and cost. For its moonshot, Cyclica is aiming to cut the 7 year drug discovery process down to 2 years by 2025. Cyclica is also working to include the large amounts of data available describing individual genetic variation to better enable personalized medicine. To further the moonshot analogy, Cyclica’s Mars-shot is to design drugs for people, not just for one protein. @Cyclica
Rare Disease Knowledge Graph
The Healx rare disease knowledge graph contains over a billion relationships enabling the derivation of novel interactions between rare diseases, their underlying biological mechanisms and drugs. This is the most comprehensive knowledge graph for rare diseases, built by integrating public and proprietary data including scientific literature, patents, clinical trials, disease symptoms, drug targets, multi-omics data and chemical structures.
Did you know that 95% of the 7,000 rare diseases in the world are without approved treatment? This affects 350 million people around the world. Healx is using AI to repurpose drugs to treat these rare diseases. By identifying existing drugs to treat rare diseases, Healx offers a cost effective way to improve the quality of life for patients. To address this huge unmet therapeutic need, the team at Healx developed Healnet. Healnet is the world’s leading AI platform for rare diseases, enabling large-scale and hypothesis-free drug discovery.
Healx can significantly reduce the time, cost and risk of rare disease drug discovery by extracting and analyzing complex information from their knowledge graph using cutting-edge machine learning methods and predictive algorithms. Biological and pharmacological rationales for their predictions are established by their expert team of pharmacologists to generate novel hypotheses for experimental validation. Healx partners with global Pharma and biotech companies to grow their pipeline using their technological expertise, either through indication expansion, shelved assets revival or identification of new therapeutic combinations. Healx is also building a robust pipeline of rare disease assets.
One of the conditions that Healx is working on is Fragile X Syndrome. Fragile X is a rare neurological disease that causes learning disabilities and cognitive impairment. Healx worked with a patient group and in about 18 months with just $100,000 managed to identify candidates of existing drugs that worked pre-clinically and that are now ready to be tested on patients in the clinic. This used to take five years and cost tens of millions – but now it only takes months and costs just $100,000. Healx is currently recruiting patients to start a trial in the US. Since these drugs are repurposed, they are already safe and on the market, so they can be accessed by patients much quicker. Healx is working to get 100 rare disease treatments to people by 2025. @healx
Minds.ai is a DeepTech company founded by deep learning purists. Most DeepTech companies focus on one domain and have only a few deep learning experts. Minds.ai focuses on deep learning and has 15 deep learning experts. The team works in tight collaboration with clients optimizing neural networks across domains including Pharma. At Minds.ai they don’t design molecules. They develop customizable AI-powered tools for drug discovery that are integrated seamlessly and with flexibility to scale easily as user base and data sets grow.
They have developed a molecular property predictor called Netrin that outperforms state-of-the-art DeepChem. Minds.ai's software libraries support deep-learning networks by communicating with graphics chips. Their library can train a neural network more quickly than leading systems. Mind.ai software was found to be 99% faster compared to other implementations. In October they contributed the low level scaling features from their training platform into TensorFlow, and co-authored a paper discussing the performance.
The team at Minds.ai is comprised of data scientists, astrophyscists, theoretical physicists, mathematicians, neuroscientists, and software engineers from all over the world. Their culture is open and their environment is distributed. Their expertise lies in cutting edge deep neural networks, such as reinforcement learning, Graph CNN’s, and Autoencoders, combined with proven expertise in massively parallel high performance computing. Minds.ai has no inherent Pharma bias as they approach every problem from a data science angle, rather than purely a biological one.
Minds.ai provides customization, workflow integration, infrastructure consulting, custom neural network development, and AI team augmentation. The company is based in Silicon Valley with offices in Bangalore and Amsterdam. @minds_ai
Insilico Medicine's mission is to extend healthy longevity through innovative AI solutions for drug discovery and aging research. The company focuses on the generation of novel drug candidates with specified molecular properties for precise disease targets. Insilico Medicine starts with molecular leads that have been specifically designed, in terms of their pharmacokinetic and pharmacodynamic properties, and therefore have a higher probability of being effective for specific disease targets.
Generative Adversarial Networks
Insilico Medicine screens potential drug candidates using GANs. GANs create synthetic datasets that are indistinguishable from real datasets by having two neural networks compete against each other. One neural network generates the data and the other compares it to a real data set in iterative cycles so that the degree of error in the synthetic data set is gradually decreased. Rather than using trial and error when looking for molecular leads, requests are made to the network to generate specific leads and leads are generated on demand.
In April, Insilico Medicine plans to relocate its headquarters from the United States to Hong Kong. The move to Hong Kong, a special administrative region of China, is the first step for Insilico Medicine in a planned expansion into China. AI is about data and China has more data than any other country. Insilico’s partners include GSK and Shanghai-based WuXi AppTec and the company has research centers in Taiwan, Russia, the UK, and Korea. @InSilicoMeds
AI for Drug Discovery Q4 Report
Our new report AI for Advanced Drug Discovery, Biomarker Development, and R&D Landscape Overview 4th Edition was published last week. In addition, we have published a special report naming the Top 100 AI Leaders in Drug Discovery. The report also includes a special guide to AI in Pharma Conferences in 2019. Deep Knowledge Analytics, a subsidiary of Deep Knowledge Ventures, regularly produces analytical reports on topics related to DeepTech, including AI in Drug Discovery and AI in Healthcare. This quarter's report is extra special because it includes two really useful and interesting supplements.
We’re Interested in Your Feedback
This article was written by Margaretta Colangelo and Dmitry Kaminskiy.
We're interested in your feedback - please leave your comments in the comments section. Click the subscribe button at the top of this article to have our articles delivered directly to you.
Margaretta Colangelo, Managing Partner at Deep Knowledge Ventures, is based in San Francisco. Margaretta serves on the Advisory Board of the AI Precision Health Institute at the University of Hawai‘i Cancer Center.
Dmitry Kaminskiy, General Partner at Deep Knowledge Ventures, is based in London. Dmitry is Managing Trustee of the Biogerontology Research Foundation.
Deep Knowledge Ventures is an investment fund focused on DeepTech. Investment sectors include AI, Precision Medicine, Longevity, and Neurotech. Deep Knowledge Ventures led Insilico Medicine’s seed funding round in 2014 and has remained a close advisor in the company’s journey towards becoming a global leader in the application of advanced AI, particularly deep learning and GANs. @DeepTech_VC