Saama Data Analytics Platform Runs On Machine Learning
By Deborah Borfitz
September 6, 2019 | In the world of clinical development, information is stored in a multitude of places, from clinical trial management systems and electronic data capture systems to third-party labs and Excel spreadsheets. Understanding if important milestones are being hit or missed requires a harmonized view of all that data in real time and in an actionable format—ideally, in the context of what users are most needing to know.
This is the playground of Saama Technologies, whose Life Science Analytics Cloud (LSAC) makes pragmatic and targeted use of machine learning to create knowledge maps specific to job roles and query preferences, says Amit Gulwadi, senior vice president of clinical innovations. The platform unifies disparate datasets to address the big and important challenges facing everyone, from study monitors to the CEO.
It’s a highly customized experience with an always-available virtual assistant named DaLIA (Deep Learning Intelligent Assistant) that has a good idea what individuals would like to do or know based on previous queries, Gulwadi says. And her answers include important parameters, such as names and locations, as well as expressions of times, quantities, monetary values, and percentages.
With the latest LSAC upgrade, DaLIA can predict users’ intentions because she has been trained on a library of over 2,500 questions, says Gulwadi. The information can be surfaced from any part of the application rather than on a certain dashboard, and a question never has to be asked twice in the same context. DaLIA remembers the interests of individual users—be it site performance statistics or adverse event rates among study participants—and their favorite questions appear at the top of the screen.
The platform’s new operational and financial risk mitigation feature uses machine learning to make on-the-fly predictions about when key performance indicators (e.g., first site activated and first and last patient enrolled) will be achieved or are in danger of being missed, he continues. “So even a small biotech company can leverage cutting-edge AI [artificial intelligence] capabilities without having to hire a single data scientist to do those calculations.”
Machine learning also underpins a new drug efficacy and patient safety analytics feature of the LSAC, reducing the time and effort required to correlate patient profiles with data variables, Gulwadi says. Current computing capabilities allow analysis of up to 50 variables simultaneously in search of outliers, in lieu of examining them manually in small batches. The feature will help researchers more quickly detect patient deviations that could spell trouble, including safety events and nonresponders, as well as identify subpopulations who are responding better, he notes.
The format of the raw data is irrelevant because the analytics are all based on CDISC standards, says Gulwadi, which are now required for regulatory submissions made in the U.S. as well as Japan.
Saama’s focus has been on biotechnology companies operating in the U.S., says Gulwadi. Pharmaceutical companies were added to the list with Saama’s recent acquisition of Comprehend Systems, which also specializes in data analytics across disparate data sources.
The LSAC roadmap includes using AI to develop persona-based insights and push the information to users via email, text or another modality, Gulwadi says. Future versions of the platform will also offer web-based monitoring and statistical analysis tools that automatically unearth risks captured in the text fields of case report forms and alert trial managers. In addition, DaLIA and the data libraries will be further enhanced.