NEW YORK (GenomeWeb) – Computational drug discovery firm TwoXar is hoping to build a business around licensing to pharmaceutical companies potential drug candidates that it identifies using proprietary algorithms that are designed to look for and explore associations between compounds and diseases.
TwoXar joins existing computational drug discovery players such as NuMedii and Cyclica, both of whom offer pharma customers proprietary platforms for matching drugs to diseases. TwoXar's approach is based on its DUMA platform, an Amazon Web Services-based drug-discovery system that uses the company's patent-pending algorithms to identify and rank high-probability drug-disease matches using a combination of public and private datasets. It helps drug developers prioritize existing candidates based on efficacy, perform targeted searches, and identify novel drugs for diseases of interest.
Recently, the Palo Alto, California-based company raised $3.4 million in a seed financing round led by Andreessen Horowitz with support from existing investors CLI Ventures and the Stanford-StartX Fund. In total, the company has raised $4.5 million from investors, and is using the funds to expand its engineering and commercial teams and to support new and existing partnerships focused on drug candidates for metabolic and neurological diseases such as Parkinson's disease, rheumatoid arthritis, type 2 diabetes, and schizophrenia, CEO and Co-founder Andrew Radin told GenomeWeb. Radin said that the company plans to add three or four hires to its current headcount of seven employees.
The company is not discussing the exact mechanism of its drug-disease association approach as it is still pursuing a patent for the technology, however Radin described it as a way of automatically constructing drug-disease interaction networks using aggregated datasets including gene expression, protein interaction networks and binding information, chemical structure, and clinical information about drug use and efficacy signals.
"The core of our technology is being able to pull in data from lots of very diverse sources [including] biological, physical structures, [and] clinical records, and ... automatically tie those very different datasets together in a manner that allows us to make predictions," Radin explained.
The system uses data from public databases maintained by the National Institutes of Health and the US Food and Drug Administration, as well as from private repositories maintained by foundations such as the Michael J Fox Foundation for Parkinson's Research. "We pull down the datasets on our platform ... we use [AWS] dynamic storage infrastructure to store the components of the datasets that we are going to access and then we actually convert them into the form that we use to ultimately to do our processing on," Radin said.
TwoXar is offering its drug discovery services primarily to pharmaceutical clients. Sample projects that the company is interested in include working with pharma companies that have existing lists of compounds and need help prioritizing candidates for preclinical studies. "Maybe they've got hundreds, even millions, of these potential drugs, [and] they can't test them all in animal studies [because] it's too expensive," Radin said. "We'll take in that set of data, process it with our system and tell them which drugs have the highest likelihood of getting an efficacious result."
TwoXar's system can also help pharma firms on the hunt for new candidates that could be developed into treatments. These clients may or may not have compound lists to share with TwoXar; instead, they might have a disease that they are interested in locating new treatment candidates for. In this scenario, TwoXar uses its algorithms to comb public and proprietary datasets for novel drug compounds that could work as treatments for the disease in question, Radin said.
In both of those sample scenarios, TwoXar runs lists of compounds through its platform. Each drug in the list is assigned an efficacy score that represents the percentage likelihood that the compound in question will treat the disease. That score is used to rank order the list of potential drug candidates from most to least efficacious. Typically, candidates that top the list will be existing treatments but interspersed in between those known treatments will be new compounds that companies could potentially pursue, according to Radin. "The actual time for the computation to produce results is just a few minutes," he said. "This is something that is replacing [existing] wet lab processes that can take anywhere from three to six years."
TwoXar's business model is based on forming revenue sharing arrangements where it licenses candidates identified by its system to pharmaceutical companies and receives a share of profits from sales of those drugs. The exact nature of these arrangements will differ from one deal to the next with the structure of each deal depending on the preferences and needs of the client in question, Radin told GenomeWeb. Deals could, for example, be based on successful completion of milestones while other deals might be based on success in a clinical study. TwoXar is currently working on early agreements with some pharma companies but Radin declined to disclose who these potential customers are.
Currently, the company is conducting early pilot projects with collaborators such as the University of Chicago, Michigan State University, and the Parkinson's Progression Markers Initiative. It has identified a number of drug candidates for various diseases that are currently being tested in preclinical studies with its collaborators.
"We are certainly not the only company that's using bioinformatics techniques in drug discovery," Radin said. "The key thing that we can do that other people can't is ... take a wider diversity of data and find the hidden associations within those datasets and do it in a way where we are not writing custom software every time we want to integrate a new dataset."
Commenting on competition from existing computational drug discovery firms, Marina Sirota, an assistant professor at University of California, San Francisco's Institute for Computational Health Sciences and scientific advisor to TwoXar, noted that there is room for multiple players in the space.
"As we build up both on the molecular and clinical side, developing computational tools to integrate and mine these datasets provide a great opportunity to cut the time that it would take to figure out different applications both for existing compounds but also in terms of identifying new drug targets," she told GenomeWeb. "There are lots of different pharma companies and I think different types of partnerships will be feasible."