How Computational Chemistry Helps Drug Discovery
Steve Jobs once said that “the biggest innovations of the twenty-first century will be the intersection of biology and technology,” and that “a new era is beginning” that he likened to the digital age that he spent most of his life working in. Numerous examples keep popping up that support his claims. You can now take someone’s DNA, the recipe for their existence, and hand it over to them on a thumb drive in digital form. We’ve even created new DNA letters and expanded the genetic alphabet. We’re using software programs to create synthetic organisms for the chemical industry that remove reliance on petroleum. We have the ability to edit genes at the germline and potentially remove hereditary diseases. And we’re now able to start modeling how things actually work in nature. One exciting area of work where biology meets technology is computational chemistry.
Simply put, computational chemistry is a branch of chemistry that uses computer simulations to assist in solving chemical problems. It’s not anything new, but the application of machine learning to computational chemistry is creating new opportunities. A paper on the topic describes this as follows:
Coupled with the maturity of GPU-accelerated computing for training deep neural networks and the exponential growth of chemical data on which to train these networks on, we anticipate that deep learning algorithms will be a valuable tool for computational chemistry.
It’s yet another application of deep learning that’s emerging and now attracting lots of funding, particularly in the area of drug discovery. Computational chemistry can accelerate the long and costly drug discovery process which can be explained by the below diagram taken from a paper on the topic published in the Encyclopedia of Nanotechnology 2015.
The trillion dollar pharmaceutical industry is currently facing a crisis of declining productivity, spending more on research each year, yet achieving fewer breakthroughs per dollar. Computational chemistry promises to decrease time-to-market and lower the overall cost of drug discovery. Consequently, venture capitalists have been pouring money into startups that are using computational chemistry for drug discovery. Let’s take a look at some of them.
Some Computational Chemistry Companies
While most startups are private companies, not all private companies are startups. We usually consider startups to be recently established companies that have taken in venture capital funding. In the case of Schrödinger, they were established way back in 1990 and have slowly been taking in funding rounds totaling $137 million so far. Their most recent funding round – an $85 million Series E round – closed in January of this year with participation from Google Ventures and the Bill & Melinda Gates Foundation which led three previous investments in Schrödinger since 2010. The well-established firm is working with multiple partner companies to advance a diverse drug discovery pipeline. They also perform extensive research in-house and offer a suite of products for drug discovery and materials science.
The fact that they closed a single funding round this year that exceeded all their past funding rounds combined shows the need to scale their business in the face of increased competition from startups like this next one.
When naming your startup, it’s always fun to take the piss out of your future marketing team by choosing a company name only a handful of people know how to pronounce. That’s what XtalPi decided to do, and if you think their company name is cryptic, just wait until you read about their technology. Founded in 2014 by a group of quantum physicists at MIT, Boston startup XtalPi has taken in $67.5 million in funding to develop “state-of-the-art crystal structure prediction (CSP, also known as polymorph prediction) technology.” A comprehensive paper on how crystal structures relate to drug discovery summarizes the application as follows:
Crystallization is one of the oldest separation and purification unit operations, and has recently contributed to significant improvements in producing higher-value products with specific properties and in building efficient manufacturing processes.
In other words, it’s all about using machine learning and predictive analytics to improve the likelihood that a drug can make it through the drawn-out approval process. The majority of their funding came in the form of a Series B that closed last Fall which followed the announcement of a strategic research collaboration with Pfizer Inc. (NYSE:PFE) in May of last year. Investors in XtalPi include Google and a large Chinese representation including Renren, Tencent, and Sequoia Capital China. Google is also an investor in our next startup as well.
No technology should be considered exciting until Softbank comes by and dumps a truckload of money onto it. That’s what happened to Relay Therapeutics late last year when they closed a $400 million Series C round led by Softbank which brought their total funding to a whopping $520 million. Founded in 2016 out of Cambridge, Massachusetts, Relay is initially focusing on developing therapeutics for oncology by using a computational chemistry approach which they describe as follows:
Many drugs are inhibitors that prevent the binding of a ligand to its active site. Relay Therapeutics’ drug discovery engine is built on advances in detecting and characterizing dynamic interactions – anywhere on a protein – not just at the active site.
Put another way, Relay Therapeutics observes the motion of pharmaceutically relevant target proteins and then predicts how they will interact with hypothetical drug molecules. They believe that “putting protein motion at the heart of drug discovery” is a complete paradigm shift for the drug discovery process.
When trying to assess the merits of a startup in any given space, it’s useful to look at how much funding they’ve raised, who they’ve raised the funding from, and who they’ve partnered with. In the case of this next startup, they’ve been securing partnerships left and right. Founded in 2012, Oxford UK startup Exscientia has raised $43.7 million in funding with more than half of that arriving this past January in the form of a $26 million Series B round. Participating in that round were two publicly traded pharma companies – Evotec and Celgene – both of whom are working with Exscientia. Just last month, Exscientia announced a three-year AI drug discovery partnership with Celgene which was said to be “the largest AI drug discovery deal done in the industry to date.” Other names they’re working with include Roche, Sanofi, and GlaxoSmithKline (GSK). Just days ago, Exscientia announced their first lead molecule to be delivered under their collaboration with GSK, targeting a key pathway for the treatment of Chronic Obstructive Pulmonary Disease (COPD).
Last but not least is a startup we took a look at a few years ago, Atomwise. In that article, we talked about the basic drug discovery process and how once there is a known protein, Atomwise can then figure out what molecule can bind to it and what will be safe. Total funding for Atomwise now sits at $51.3 million after they raised an additional $45 million in the form of a Series A round that closed in March of last year with participation that included Monsanto, Chinese firms Baidu and Tencent, and notable venture capital firms that included Khosla and DFJ. Atomwise claims their deep convolutional neural network, AtomNet, screens between 10 and 20 million compounds a day. If you want to know more, here’s a paper that describes their technology (click the below image to access the entire paper).
Just this past January, Atomnet received a vote of confidence from Charles River (CRL), a $7 billion firm that offers a variety of pre-clinical and clinical laboratory services to the pharmaceutical industry. Both companies entered into a strategic relationship where Atomwise will “support hit discovery, hit-to-lead, and lead optimization efforts; while receiving technology access fees, milestone-based payments and royalties from clients,” a number that could exceed $2.4 billion according to the press release. Clients can expect to see “an exponential reduction in the cost and risk of drug discovery.”
It’s important to distinguish between “computational drug discovery” and “computational chemistry.” There are now upwards of 120 startups out there working on various aspects of “computational drug discovery.” That is, they’re using new technologies like machine learning to improve the entire drug discovery process. “Computational chemistry” also falls under this label, but specifically refers to the use of computational methods to optimize a molecule once a target has been identified. All of the startups we’ve talked about have taken in sizable funding rounds in the last year and four of them have raised in the last six months. An industry expert we spoke to speculated that this rush of capital to computational chemistry startups is because they’re optimizing a step in the drug discovery process that’s easy to understand. In our next article on this topic, we’ll take a look at some of the progress being made in the broader “computational drug discovery” space.