Machine Learning in Pharma: Mindbreeze InSpire

Where are pharma companies investing their windfall profits? The companies are evolving in many ways, including stepping outside drug development to tryout digital treatments and using lessons from the COVID-19 pandemic to rethink production and clinical trials.

I spoke recently to Daniel Fallmann, founder and CEO of Mindbreeze, about what pharma companies are doing with their data mining and machine learning platform.

Surfacing hidden information

Mindbreeze’s service, Mindbreeze InSpire, slurps up pharma companies’ data and serves it up to knowledge workers, who are looking for patterns that can help them identify new relationships and ultimately derive clinical or business insights. For instance, some of Mindbreeze InSpire’s main users are reviewers of the companies’ drug submissions. These reviewers employ Mindbreeze InSpire to scan all sorts of information maintained by the company—studies, specialist catalogs, information about trials, competitor intelligence, laboratory findings, market reports, even email and other forms of internal communication—to find important facts that might indicate safety and efficacy. Mindbreeze InSpire can narrow searches to particular populations, such as to see whether Latinx males over the age of 50 have an unusual reaction to a medication.

Another example use is a researcher who can turn up details about a specific drug ingredient in less than a minute.

One of the most intriguing uses for Mindbreeze InSpire is to discover specific types of human expertise within an organization. In this way, AI doesn’t replace human thinking but allows humans to work together more effectively.

Fallmann mentioned, as an example, a large U.S. organization employing 200,000 people that needed, after COVID-19 struck, to quickly find internal experts with very specific skills. They tried direct approaches such as issuing self-assessments to their staff and scraping LinkedIn profiles, but none of these worked. People weren’t identifying the skills urgently needed, either because they didn’t value having these skills or wanted to spend their time on other activities.

Mindbreeze InSpire solved the company’s problem by analyzing the content of staff’s everyday communications, as well as research papers on which the staff played a role, possibly years before. The more often a keyword was mentioned in these inputs, the higher the relevancy of the employee.

Fallmann says, “Like all companies, pharmaceutical companies have information that could benefit them, but that they aren’t using because they don’t know it exists.”

Architecture and activities

Mindbreeze InSpire creates a knowledge graph for a company, using graph database software with an inverted graph index. To retrieve information, the service employs connectors appropriate to each type of input (databases, web services, documentation, etc.). The software recognizes object definitions when the input is structured, so as to extract the relevant information semantically from each input. Security is also respected; knowledge workers can’t see fields to which they’re denied access by the original sources.

Some sources are polled regularly for new material, whereas others (such as databases) have triggers that push new data to Mindbreeze InSpire’s connector. The power of Mindbreeze Inspire is its ability to combine information from diverse sources, but it recognizes when measurement scales or other traits are incompatible, and doesn’t try to mash together data sets that would produce erroneous results. Mindbreeze has worked hard to make updates to the index quick so they can be done continuously.

The front end provides graphical tools that let non-programmers build models. Mindbreeze therefore joins a trend in modern computing that helps knowledge workers create customized tools for themselves without involving the professional IT team.

Although their analytics engine started as a software service, Mindbreeze soon realized that they had to provide their partners with a hardware appliance. Their customers were trying to run their compute-intensive machine learning on standard computer hardware. As every machine learning enthusiast knows, generic chips are not enough—that’s why machine learning didn’t take off until researchers discovered they could run their algorithms on off-the-shelf GPUs, which are now a standard offering in every cloud vendor. The Mindbreeze appliance runs on Dell hardware with NVDIA GPUs.

Mindbreeze was founded in Austria in 2005. The group employes about 150 engineers at their Linz development headquarters and 350 staff total around the world. Their headquarters in Chicago serves the U.S. market. Fallmann says they do well in the U.S. because “organizations here love commercial off the shelf products and like leading edge technology,” which is well suited to their AI offerings. Indeed, the way AI is being offered up by Mindbreeze, it could become the bedrock for business decision-making.

About the author

Andy Oram

Andy is a writer and editor in the computer field. His editorial projects have ranged from a legal guide covering intellectual property to a graphic novel about teenage hackers. A correspondent for Healthcare IT Today, Andy also writes often on policy issues related to the Internet and on trends affecting technical innovation and its effects on society. Print publications where his work has appeared include The Economist, Communications of the ACM, Copyright World, the Journal of Information Technology & Politics, Vanguardia Dossier, and Internet Law and Business. Conferences where he has presented talks include O'Reilly's Open Source Convention, FISL (Brazil), FOSDEM (Brussels), DebConf, and LibrePlanet. Andy participates in the Association for Computing Machinery's policy organization, named USTPC, and is on the editorial board of the Linux Professional Institute.

   

Categories