7 Companies Using AI for Drug Discovery
Drug development is complicated business. Medicinal chemists, many who have spent nearly a decade training in the halls of academia, are tasked with painstakingly synthesizing each and every molecule that could potentially have a therapeutic effect. These compounds need to then be screened in cells just to make sure they’re not just straight-up toxic, with the small possibility that they have some statistically significant effect on a model of a diseased cell.
After that, these drug candidates are tested in animals to get a little closer to the real deal – the clinical trials. Only 1 in 1000 potential drugs ever reach human testing. A harrowing gauntlet that has drug candidates tested through four phases with real human patients, this is where the wheat is separated from the chaff by the FDA, with only 1 in 5 drug candidates making it through. The entire process takes an average of 12 years for experimental drugs to start from the lab bench and wind up in the medicine cabinet. That means up to $2 billion in costs and endless failures along the way. Big pharma is all about going big or going home.
Artificial intelligence and machine learning promise to do away with such inefficiencies. Investment in the use of AI for drug design and discovery has received the most out of all private AI investments in 2020, with more than $13.8 billion put into this arena. That’s because drug design and discovery are already a costly venture and everyone in the pharmaceutical industry needs all the help they can get. More accurate predictions and lower lead times translate to more green down the road.
Here are 7 companies that are employing the power of AI to discover the cure for
erectile dysfunction cancer and other serious human maladies.
AI Platforms for Drug Design From First Principles
Founded in 2014, New York-based Roivant Sciences has received a massive amount of funding, to the tune of $1.9 billion after bringing in $40 million from a Series B that closed in 2020. The company is accelerating drug development for late-stage drug candidates by building subsidiaries known as Vants. These are subsidiary companies, all of which are cleverly given names that end in ‘vant’. Roivant currently has over 40 drugs in development across its family companies. Valuation in this unicorn biotech umbrella group has swelled to a jaw-dropping $7 billion.
The company recently acquired Silicon Therapeutics in February 2021 for a whopping $450 million to augment its computational drug design capabilities. Silicon Therapeutics is touting a physics-driven approach to drug design and development. The strategy boils down to using first principles from physics to determine if a drug candidate exhibits any biological effect, rather than applying assumptions drawn from empirical observations in chemistry and biology about the drug development process.
This method relies on a mathematical model of the subatomic universe, called a force field, that governs how molecules should interact with each other. One of the most important interactions is the fit between drugs and cell receptors that trigger cell signaling cascades, which is one way therapeutic effects actually occur. More accurate models allow drug design companies, like Silicon Therapeutics, to get more potential drug hits in real life. Quantum mechanics, statistical thermodynamics, and molecular simulations are all packaged in a software platform that’s accelerated using a supercomputing cluster. The results of these molecular simulations are then used to drive decisions at the wet benchtop (in other words, a traditional laboratory).
(Editor’s Note: The last time we looked at “the vants” was back in 2017 when we told A Story About Roivant, Axovant, and Myovant. It’s nearly impossible to keep up with this family of drug development companies, so we’ll keep an eye on the exit instead. Should any of these become available to retail investors via IPO, we’ll vet them to see if it merits an article for Nanalyze Premium subscribers.)
Founded in 2018, insitro is a data-driven drug discovery and development company headquartered in San Francisco. The company has raised $743 million after a Series C round that pulled in a massive $400 million in March 2021, drawing investments from the likes of Andreessen Horowitz and the Canada Pension Plan Investment Board. insitro was founded by Dr. Daphne Koller, a professor of computer science at Stanford University, recipient of the MacArthur Fellowship (also known as the Genius Grant), and co-founder of Coursera alongside Andrew Ng. We talked about insitro landing a three-year deal with Gilead (GILD) back in 2019.
Dr. Daphne Koller and her team are addressing one of the sticking points that’s created challenges in bridging the gap between computational biology and machine learning – data. Existing biological data is limited or too low quality to be machine-readable. insitro, a portmanteau of in vitro (cell studies) and in silico (computational studies), is building out a pipeline of experimental data gathered from pluripotent stem cell studies to create a giant array of data sets based on real biological systems. Using this data bank, the company will use machine learning to generate predictions on how healthy and diseased cells behave inside a human being, and what will happen if drug candidates are introduced into a simulated model of a diseased system. That certainly deserves the moniker of genius.
Based in Cambridge, Massachusetts, XtalPi was founded by quantum physicists from MIT in 2014 and has since pulled in $386.4 million from investors such as Sequoia Capital China and Google (GOOG). The company completed a Series C funding round led by SoftBank (SFTBY) in September 2020 for $318.8 million. We took a look at XtalPi back in 2019, and despite still not having a pronounceable name, it looks like the company is impressing the investor crowd with its technology.
XtalPi has built out its Intelligence Digital Drug Discovery and Development (ID4) platform, a machine learning platform that predicts the crystal structure and properties of small-molecule drug candidates using principles from quantum mechanics and solid-state crystallography. The company partnered with Pfizer (PFE) back in 2018 to build out its AI-powered platform.
Update 08/12/2021: XtalPi has raised $400 million in Series D funding to accelerate drug discovery. This brings the company’s total funding to $786.4 million to date.
AI Platforms for Drug Design of Protein Therapeutics
Founded in 2011, Washington-based AbSci is a bioengineering startup that’s raised $238 million from names like Fidelity, Merck (MRK), completing a recent venture round for an additional $125 million back in March 2021. The company is building its Protein Printing Platform Technology, which is composed of two parts – SoluPro, a synthetic biology platform that generates protein-based drug candidates from genetically engineered E. coli bacteria, and SoluPure, a purification platform that rapidly extracts and purifies the target proteins.
The platform relies on AI to help produce and screen for novel drug candidates. To continue powering its data-heavy tech, the company acquired the Devonium Engine machine learning platform in January 2021. AbSci’s technology can shorten preclinical timelines and has even developed a COVID-19 antibody cell line in 10 days. With super futuristic-sounding machine learning software now under its belt, how can AbSci go wrong?
AI Platforms for Drug Design using Chemical Principles
Founded in 2012, Exscientia is a pharmatech startup headquartered in Oxford. After closing a Series C round that ended in March 2021 for $40 million, the company has raised a grand total of $147.9 million from Bristol-Meyer Squibb (BMY), Evotec (EVT.DE), the Bill & Melinda Gates Foundation, and others. Exscientia has developed an AI-powered design platform known as Centaur Chemist, which relies on machine learning methods to search for potential drugs in the same manner as pharmaceutical chemists, only faster.
The process of drug discovery is similar to chess, in that there are only so many possible moves in the field of medicinal chemistry, but the number of possible variations is too staggering for a computer to apply brute force methods. Instead, the AI platform learns the limited possible approaches used in traditional drug discovery from human chemists, such as retrosynthetic analysis, and improves with each sequential iteration. The platform informs drug discovery experts on the next move, who provide additional rules or constraints if such moves are not chemically possible. The joint effort between AI and humans has shortened the average time of bringing drugs to phase 1 from five years to only 12 months. Not quite the man-machine cybernetic hybrid we all hoped for, but it works.
The company has recently announced its first AI-designed molecule for treating cancer using the body’s own immune system, co-developed with Evotec (EVT.DE), will be entering human clinical trials. This comes following its 2020 announcement of bringing an AI-designed molecule for treating obsessive compulsive disorder (OCD) to phase 1 clinical trials in partnership with Japanese collaborator Sumitomo Dainippon Pharma (DNPUF).
Founded in 2019, PostEra is another San Francisco-based company that straddles between the disciplines of medicinal chemistry and machine learning. The startup has received $2.5 million in disclosed funding after a series of seed rounds that included Y Combinator and is currently partnered with Pfizer (PFE) to advance machine learning for drug discovery. PostEra’s platform relies on the design strategies and rules of chemical synthesis used by medicinal and organic chemists to synthesize drug candidates.
Pfizer plans to provide the startup with mountains of data to power PostEra’s platform. The company also recently led the COVID Moonshot initiative to discover chemical compounds that could serve as antiviral agents against COVID-19 using its platform. The initiative crowdsourced 14,000 submissions from medicinal chemists, which led to four drug candidates being tested in animal studies. Who knew that an army of
out-of-work chemists with a side hobby could do so much good?
AI Platforms for Discovering Natural Products
Many therapeutics used today are still derived from natural products. For example, the active ingredient of aspirin, acetylsalicylic acid, is a derivative of a compound found in willow bark, and the anticancer agent taxol is manufactured from a compound produced by the yew tree. Even the $2 billion blockbuster drug, lovastatin, is chemically identical to a compound found in red yeast rice, a staple food in which rice is fermented with an edible mold.
San Francisco-based Brightseed is a biotech startup founded in 2017 to discover new health-promoting molecules from the natural world. The company has brought in $27 million from a Series A funding round in September 2020 to do just that. Brightseed has designed an AI discovery platform known as Forager to scour the world of edible plants and food for natural compounds that have an impact on human health. Known as phytonutrients, many of these compounds are used as functional ingredients in food but their effects on health are still not fully known.
So far, the company has analyzed over 700,000 bioactive compounds and has discovered one natural food ingredient that can support the health of the liver by managing fatty liver disease, a condition associated with obesity. Forager will be used to expand the known set of food-based nutraceuticals. Brightseed will partner with food companies, including Danone North America (BN.PA), to help them develop products that support human health as well as bring its own functional ingredients to market.
The future looks bright for the marriage between machine learning and drug design and discovery. With so much data available from decades of research generated by the pharmaceutical industry and academia, with even more machine-readable data being produced in the upcoming years, the day that AI can successfully design, synthesize, and even market its own drugs is coming soon.
But before AI learns how to build lab robots, computers will still need human hands to make sure those drugs get made, screened, and sold. That means plenty of PhD-level synthetic chemists will be employable at the benchtop rather than as an adjunct professor at a university. At least for now.
There’s one AI drug discovery stock we’re holding in our disruptive tech stock portfolio. Become a Nanalyze Premium annual subscriber and see all 32 of our holdings today.