Representatives from Stanford University spinoff Ayasdi told BioInform this week that the firm’s query-free approach to data analysis could reveal hidden insights in data that could result in better drugs and more targeted therapies, while also setting it apart from competing software vendors in the life sciences space.
Ayasdi’s platform, dubbed Ayasdi Iris, is based on a series of machine learning algorithms that search for patterns or anomalies in data and display them as topological networks or shapes.
With Ayasdi’s approach, users don’t need to formulate specific research questions they’d like answers to prior to beginning the analysis of their data, Pek Lum, Ayasdi’s vice president of products and solutions, told BioInform.
Instead, “you [let] our platform tell you where you should start your analysis” and then drill down and do “more structured or semi-structured or supervised analysis on the data,” Jeff Yoshimura, Ayasdi’s vice president for marketing, told BioInform. This allows users to explore multidimensional datasets without an existing bias, he said.
Earlier this month, the Palo Alto, Calif.-based firm announced that it received $10.25 million in a Series A funding round led by Khosla Ventures with additional capital from FloodGate.
Yoshimura said that the 30-person firm will use its new funds to hire more engineers and data scientists and to further develop Ayasdi Iris, its software-as-a-service data analysis platform.
Ayasdi was founded in 2008 by three Stanford mathematicians to commercialize software they’d developed based on topological data analysis — a mathematical approach to analysis that looks for shapes or patterns in data. Their work at Stanford was supported by funds from the US Defense Advanced Research Projects Agency and the National Science Foundation.
For the first few years of its existence, the company’s founders stayed under the radar, according to Yoshimura, and worked on developing the platform, building a front-end application, and raising a first round of venture capital to support their fledgling business.
Then in early 2011, the company launched the first version of Ayasdi Iris, which they initially sold to clients in the pharmaceutical industry and government organizations such as the US Food and Drug Administration, Lum told BioInform. A second version of the platform was launched later that same year that was marketed more broadly to pharma and biotechnology companies as well as academic and research institutions, she said.
Customers are charged an annual recurring subscription fee that changes based on the number of users and the type of datasets to be analyzed, Yoshimura said. Ayasdi has a different pricing scheme for academic groups, who also receive a yearly license to the software.
Customers can choose to use Ayasdi’s servers for their computations or they can use an internal cloud, Lum said.
Analyze First, Ask Later
Lum said that one of Ayasdi’s main selling points is that its customers don’t need to have any coding skills or expert knowledge of tools such as SAS, Matlab, or R in order to use the platform.
All they have to do is upload their data to the system and Iris automatically generates networks that reveal the relationships between points in the data that may not appear if targeted queries are used as the search criteria, Lum explained.
Identifying these patterns lets researchers ask questions that they might not have thought of otherwise and then analyze the results using a series of downstream statistical packages that Ayasdi includes in its offering, she said. Alternatively, customers can use the company’s application programming interface to connect third-party tools and data services they’d like to use.
To illustrate Iris’s analysis efficacy, Lum told BioInform that Ayasdi researchers used the system to analyze gene expression data from biopsied tumors of breast cancer patients. The data for this particular study came from the Netherlands Cancer Institute’s Antoni van Leeuwenhoek Hospital.
Lum said that the researchers were able to identify subpopulations of patients within the data whose treatment response to specific therapies was known and additional subgroups of patients who would likely benefit from other kinds of targeted therapies.
For instance, the team found two subgroups of survivors who each had molecular signatures that were distinct from other survivors in the group, Lum said. As a follow-up to the analysis, researchers might want to explore which genes or pathways might be responsible for the survival of patients in each of these smaller groups, she said.
In addition to exploring their own data, users also have access to communal resources such as the Cancer Genome Atlas through the platform, so that they can look for new patterns in the public data that conventional query-based analysis methods may have missed, she said.
The company believes that a fresh look at these existing data repositories alongside analysis of new datasets could result in new treatments and more targeted therapies for diseases such as cancer.
While Ayasdi is focusing first on the life sciences, the company is also peddling its product in other markets. These include oil and gas, financial services, the public sector, manufacturing, retail, telecommunications, and sports.
Ayasdi’s clients in these areas include the Defense Advanced Research Projects Agency, which has used Iris in some national security-related projects.
On the life sciences front, besides the FDA, Ayasdi’s customer base includes Merck, the Mount Sinai Medical Center, Second Genome, BN ImmunoTherapeutics, the United States Department of Agriculture, and the US Centers for Disease Control and Prevention, among others.
The USDA, for example, used Ayasdi to analyze data from a strain of Escherichia coli. Specifically, USDA researchers reconstructed the relationship structure of E. coli O157:H7 survival in soil samples from California and Arizona, according to Ayasdi. The analysis revealed a complex interaction between the E. coli strain, soil microbiota, and soil properties that impacted the survival of the pathogen.
Meanwhile, researchers at Mount Sinai are using the platform to analyze genetic data from multiple disease conditions such as cancer in order to identify more effective treatments. And at UCSF, researchers have used Iris to analyze data from traumatic brain injury cases to identify groups of patients who are likely to benefit from specific treatments.