Updated: 10.01.2019. Newly added content is marked in the text with "Update" sign.
The idea of using artificial intelligence (AI) to accelerate drug discovery process and boost a success rate of pharmaceutical research programs has inspired a surge of activity in this area over the last several years. In 2018, things are getting even “hotter” with the increase in the amount of partnerships, investments and other important events, summarized and grouped below into “mini-trends”.
Illustration by Andrii Buvailo
1. Venture capital is pouring into AI-driven drug discovery startups
This year has been marked by an impressive number of fundraising deals among AI-driven drug discovery startups -- a clear indication of the “AI for drug discovery” space gaining some serious attractiveness for venture capitalists.
So far, a London-based BenevolentAI appears to be a leader of the year in terms of fundraising -- in April they closed a $115 M round, reaching a staggering $2 billion valuation mark. While met with certain degree of skepticism, this news and the current pace of research activity by the company undoubtedly puts BenevolentAI in a very strong position among competitors.
Atomwise, which was founded in 2012 and pioneered the use of deep neural networks for structure-based drug design, raised $45 M round A investment to advance its AI-driven drug discovery technology AtomNet. The company says it screens 10 million small molecules each day and uses AtomNet, which is utilizing deep learning algorithms, to analyze molecules and predict their potency as medications, toxicity, and side effects.
A quite unique company on the list -- a US-based Insilico Medicine, which is the only one startup among its closest competitors which develops a “full-stack” artificial intelligence system based on generative adversarial networks (GANs), allowing for an “end-to-end” drug discovery process -- from basic biological modeling and biomarker development, to hit-molecule generation, lead optimization and pre-clinical validation of drug-candidates. In June, Insilico Medicine received an undisclosed amount of strategic investment from WuXi AppTec, bringing totally raised capital up to $20 M (according to Crunchbase).
Notably, just a month later, WuXi AppTec participated in a $32 M investment round for another AI-driven startup -- Verge Genomics. The latter uses machine learning and AI to develop therapeutics against Alzheimer's and Parkinson’s disease. Verge is also actively growing its database of patient genomic data -- allegedly, the company possesses one of the industry’s largest resources in this therapeutic area.
New York - Paris based Owkin, founded in 2016 to apply machine learning for optimizing drug discovery process via better comprehending the overabundant biological data, raised its Round A of $11M in January to scale its technology platform Owkin Socrates. The platform can integrate molecular and imaging libraries with patient data to reveal patterns of biomarkers causing a disease, and the company is applying transfer learning to improve model performance where properly labeled data is scarce.
Founded in 2014 by a group of quantum physicists at MIT, XtalPi is a U.S.- China biotech firm which has raised a Series B round of $15M in January from several investors, including Google and Sequoia China among the others. The company is claiming that it can quickly and accurately predict numerous important characteristics of small-molecule drugs and solid forms by combining artificial intelligence, quantum physics, and high-performance cloud computing. Using this sophisticated interplay of technologies, the company will be able to provide “time-saving insights into the safety, stability, and efficacy of drug candidates”.
Later this year, Google also co-invested in BenchSci -- a smart platform for AI-powered search for biological products. The round totalled $8 million from several investors.
Engine Biosciences is a San Francisco and Singapore based biotech firm, which announced a $10 M funding round to advance its AI-based platform for drug discovery, development of combination therapeutics, and cellular reprogramming. The company’s technology allows researchers and drug developers to reveal gene interactions and biological networks, and provide test therapies specifically targeting genetic interactions. The company’s AI platform can assist in target discovery, drug repurposing, and analysis for precision medicine applications.
Other notable investments in 2018 include: TwoXAR ($10 M), ReviveMed ($1,5 M), GTN ($2.8 M) etc.
2. Drug makers continue hunt for external AI-partnerships...
In 2018, pharmaceutical companies show continuous interest in partnering with emerging AI-driven startups -- to leverage the power of algorithms for boosting own drug discovery programs. Below is a list of some of the notable drug design collaborations of this kind:
The last month of this fruitful year was marked by a new research collaboration between German pharmaceutical giant Merck and a Canadian AI-driven company Cyclica. The parties agreed that Merck will use Cyclica’s proprietary AI-driven cloud based in silico proteome screening platform Ligand Express® to clarify mechanisms of action for a number of Merck’s small molecule candidates, evaluate their safety profiles and uncover additional therapeutic applications.
In November, Bayer established a multi-phase research collaboration with Toronto-based drug discovery company Cyclica to utilize its multifaceted AI-driven discovery platform for a broad range of research tasks. In the framework of this collaboration, Cyclica will provide its cloud-based proteome screening platform Ligand Express® to study the off-target profiles of small molecules, and apply its first-in-class Differential Drug Design (DDD) technology for multi-targeted drug design. Furthermore, it will apply its AI technology to build state-of-the-art predictive models for pharmacokinetic properties.
In September, Pfizer entered into an evaluation agreement with Atomwise -- now the AI-developing startup will need to identify promising drug candidates for up to three proteins of choice by Pfizer.
Just a couple of months earlier, Pfizer partnered with another AI-driven startup XtalPi to develop a drug discovery software platform, which would utilize XtalPi’s expertise in computational physics and artificial intelligence. The platform is to be applied for accurate molecular modeling of drug-like small molecules.
Bristol-Myers Squibb entered a multi-target research collaboration agreement with Sirenas, a biotech company applying machine-learning based computational approaches to discover therapeutics derived from the global microbiome, to apply its proprietary drug discovery platform against a series of undisclosed but challenging therapeutic targets. The research collaboration leverages Sirenas' expertise in applying its proprietary data mining technology ATLANTIS™ to identify potential drug candidates among Sirenas' proprietary chemical library isolated from global microbiome collections. It is important to note another area of Sirenas’ expertise -- state-of-the-art organic synthesis, which makes it possible for the company to deliver not only computational predictions but also chemical compounds with unusual nature-inspired scaffolds.
In May 2018, Boehringer Ingelheim partnered with Bactevo to apply their "Totally Integrated Medicines Engine" for identifying novel small molecule drug candidates.
In May, GlaxoSmithKline (GSK) has formed a drug design collaboration with Cloud Pharmaceuticals, an AI-drive drug discovery company, to develop a series of small molecules against biological targets specified by GSK.
3.… but also expand internal AI capabilities
On the one hand, pharmaceutical companies are increasingly hiring AI-startups to explore opportunities, but on the other hand, they are equally active in growing internal AI expertise and shaping digital infrastructures for more efficient data usage.
Recently, Novartis announced the completion of the first phase of a company’s digital transformation strategy focusing on big data, digital infrastructure and artificial intelligence. The first phase was internal program called STRIDE, and it included the launch of several important IT infrastructure systems for document management, internal investigation, high performance computing, clinical trial management and other tasks.
The next phase of Novartis’s digital transformation is to implement a predictive analytics platform, driven by machine learning algorithms, to support clinical trial operations. This will be done in the framework of Nerve Live initiative, and in collaboration with US machine learning company QuantumBlack.
Finally, there are plans for the third big future project -- Data 42 -- the one bringing all of Novartis’ data sets together to be able to query any data in a centralized mannar. This is, certainly, a major prerequisite for the AI-driven transformation in the company.
Similarly, pretty much every global drug maker -- Pfizer, AstraZeneca, Eli Lilly, Merck, GSK and others -- are taking internal restructuring measures to get prepared for the digital transformation of pharmaceutical research and adoption of artificial intelligence for drug discovery and development.
4. Hunting for “big data”
It is becoming obvious, that the key enablement factor of the future “AI-driven revolution” in pharmaceutical research is data. Without the access to diverse, interdisciplinary, quality, and properly curated big data, a transformative impact of AI technology can not be fully realized. In this context, it is important to see how companies are moving in the direction of data-centric research paradigm.
GSK and 23andMe
In July, GSK has invested $300 million in 23andMe, a Silicon Valley gene testing company, backed by Google. This deal opens a door for GSK to access a vast DNA database, providing information about the relations between genes and diseases. 23andMe has more than 5 million customers, the majority of whom opted in to allow their data being included in research programs.
Datavant and Verge Genomics
Datavant, a young US-based AI-driven startup, is focused on organizing and structuring healthcare data for deriving actionable insights for the design and interpretation of clinical trials. In the beginning of January, it announced a strategic alliance with Verge Genomics, a company using artificial intelligence to discover and develop new therapeutics. The newly formed partnership aims at unlocking the value of pharmaceutical datasets in a possession of Datavant -- clinical trial data, claims, pharmacy history, electronic health records and genomics data on patients.-- to accelerate discovery and development of new medications.
So far, Datavant has two more partnerships besides Verge -- with Duke Clinical Research Institute (DCRI), Global Genomics Group (G3) -- all aiming at combining drug discovery expertise, biological big data, and novel data analytical technologies, such as AI, to boost innovation in the field of pharmaceutical research.
5. Moving towards integrated research platforms
In the light of the above trends (focus on AI and big data), a logical consequence is the pharmaceutical research industry moving towards platform-based models of cooperation and doing research. Platforms are digital infrastructures, connecting the dots between different types of activities, research areas, operation modes, and data flows. Platforms, or “super-platforms” are widespread in finance, consumer e-commerce, and other industries, but this is still a new phenomenon for the pharmaceutical research. Several events in 2018 are quite illustrative here:
Merck, Accenture, and AWS
It was announced that Merck and Accenture are working with Amazon Web Services to create a cloud-based platform that would embrace collaborators across various sectors of the life sciences industry. This analytics platform will be built using open application programming interfaces (APIs), and will facilitate a collaborative environment to accelerate early drug discovery efforts. It will not only make it easier for researchers to aggregate, access and analyze interdisciplinary data, but will also lower barriers to market entry for novel value providers -- app developers, data scientists, content and data suppliers etc.
Google and WuXi NextCODE
In March, WuXi NextCODE announced a partnership with Google to integrate its massively scalable genomics database management system and research apps in Google Cloud Platform. In turn, such tools as Google Cloud BigQuery, and DeepVariant will be integrated with WuXi NextCODE’s capabilities. The two companies will also work on additional tools and APIs to empower the global genomics community.
6. Organizations join forces to adopt AI for drug discovery
One of the important elements of a mature industrial ecosystem -- the presence of specialized consortia and associations, whose goal is to facilitate interaction between members of the community, set industry standards and reveal best practises, educate general public about the topic, and lobby important changes to government regulations.
The pharmaceutical research industry is in its early days of a widespread adoption of artificial intelligence for drug discovery, so the emerging ecosystem of AI practitioners in this space is only beginning to grow. However, a number of important steps towards the creation of industrial alliances has already been made recently:
In May 2018, MIT has formed a powerful industry-academia consortium, the Machine Learning for Pharmaceutical Discovery and Synthesis (MLPDS), which already includes some of the leading players in the pharmaceutical field: Amgen, BASF, Bayer, Eli Lilly, Novartis, Pfizer, Sunovion, and WuXi. Being headquartered in Cambridge, MA, one of the global centers for biopharmaceutical innovation, the newly formed consortium allows for close cooperation between partners (a lot of them have presence in Cambridge), and the creation of a center for artificial intelligence (AI) use in pharmaceutical research.
Another important consortium, The Accelerating Therapeutics for Opportunities in Medicine (ATOM), has been formed at the end of the last year by its founding partners -- GSK, Lawrence Livermore National Laboratory, Frederick National Laboratory for Cancer Research, and the University of California, San Francisco -- with funding support under the 21st Century Cures Act. While the ATOM’s mission includes a broad range of activities to facilitate efficient drug discovery in the field of oncology, some of the central tasks are focused on advancing artificial intelligence adoption by pharma players and democratising access to research big data. In April 2018, Numerate, one of the leading AI-developers from drug discovery, expressed its intentions to join the consortium.
Finally, September was marked by an important milestone -- the announcement of mission and launch activities of a global Alliance for Artificial Intelligence in Healthcare (AAIH), which is to become a leading international organization for advancing artificial intelligence innovations in Drug Discovery, Clinical Research, Diagnostics, Precision Medicine and other key areas of pharmaceutical research and healthcare.
7. Creating and industry benchmark for comparing machine learning models [Update]
Having a standardized set of metrics and datasets for assessing and comparing a wide variety of available and novel machine learning models is essential for creating and maintaining industry’s best practises.
MOSES (Molecular Sets)
A recent move in this direction has been made by a group of scientists from AI-driven drug discovery company Insilico Medicine, in collaboration with a distributed synthetic data platform for deep learning -- Neuromation, and Alán Aspuru-Guzik's research group at the University of Toronto, who launched an open research platform MOSES (Molecular Sets), described in the paper "Molecular Sets (MOSES): A Benchmarking Platform for Molecular Generation Models". The source code and datasets for the platform are all available at GitHub.
The platform is supposed to play a similar role in boosting AI-driven drug discovery, as ImageNet played in advancing deep learning for imaging data. MOSES is open for researchers and organizations to contribute their datasets and models to extend the benchmarking platform.
The above post summarizes very briefly some of the aspects of how artificial intelligence technologies and big data are starting to play the central role in the pharmaceutical research. To get a more comprehensive view on the subject, please, contact us via firstname.lastname@example.org to request quotation for the analytical report “The Opportunities and Challenges of Adopting AI in Drug Discovery and Pharmaceutical R&D”.
Please, subscribe to our newsletter to get fresh market analytics insights directly into your inbox -- we will rarely bother you more than once per month.