EVENTS | VIEW CALENDAR
Accelerating AI adoption
BOSTON & LONDON—Because the U.S. Food and Drug Administration (FDA) believes that artificial intelligence (AI) and machine learning (ML) can transform the delivery of healthcare and wants to create an AI regulatory framework, The Pistoia Alliance—a global non-profit organization of life-science companies, technology and service providers, publishers, and academic groups attempting lower barriers to innovation in life sciences R&D—conducted a survey of barriers that could hamper the potential of AI.
Conceived in 2007 and incorporated in 2009 by representatives of AstraZeneca, GlaxoSmithKline, Novartis and Pfizer who met at a conference in Pistoia, Italy, The Pistoia Alliance undertakes projects that transform R&D through pre-competitive collaboration. It overcomes common R&D obstacles by identifying the root causes, developing standards and best practices, sharing pre-competitive data and knowledge and implementing technology pilots. More than 150 member companies collaborate on projects that generate significant value for the worldwide life-sciences R&D community.
The research, which surveyed 190 life-sciences professionals across the U.S. and Europe via a series of webinars between October 2018 and February 2019, showed that access to data (52 percent) and lack of skills (44 percent) are the biggest barriers to the adoption of AI and ML. In 2017, 24 percent of scientists queried said that access to data was the biggest challenge to AI adoption, while 30 percent cited lack of skills. Still, AI usage in the life sciences has increased during that time, with 70 percent of respondents using AI, up from 44 percent in 2017. Dr. Steve Arlington, president of The Pistoia Alliance, believes that much can be done to address the issues of data quality, access to data and lack of skills.
“Data quality is a problem the life-science industry can take immediate steps to address,” Arlington said. “The industry must ensure that data is compliant with the ‘FAIR’ principles: Findable, Accessible, Interoperable and Reusable. The FAIR principles were published in 2016 with the goal of emphasising machine-actionability (making data interoperable and re-usable), because humans increasingly need computational support to manage the volume and complexity of data.”
He explained that The Pistoia Alliance is working with its members on how they can adopt FAIR. Companies need to adopt industry-wide data standards, one example being the Unified Data Model (UDM). This project will create and publish an open and freely available data format for storage and exchange of experimental information about compound synthesis and biological testing.
“When it comes to skills, upskilling those already in the industry will be a key factor in improving AI, as well as altering job-seekers’ impressions to attract skilled data scientists to pharma roles,” Arlington added. “Today, life-sciences companies don’t find it easy to attract digital natives; there is often a pay discrepancy between the science and tech industries, and pharma has not typically been recognized as a sector that is leading from the front when it comes to digital innovation.”
Arlington believes that the industry needs to build closer ties with academic institutions to design courses that teach the skills that will accelerate AI adoption, commenting that “There is an abundance of data streams in R&D today—such as real-world evidence, clinical trials data, wearable and genomic data—which could have real value in drug discovery, but only as long as we have the skills to analyze it. By working with academia and with educators to highlight these innovative opportunities, the life-science industry can ensure that it is fostering the next generation of data scientists. AI has the potential to make a real difference in life sciences, particularly when we look at how data can be used for good, such as in the drug repurposing datathon we ran with Elsevier earlier this year.”
“Collaboration is key to implementing data standards, and a project like the Unified Data Model (UDM) shows this power of collaboration in action,” he remarked. “The UDM format was originally donated to The Pistoia Alliance by global analytics company, Elsevier, and is now being developed under the stewardship of the Alliance’s project team, with funding from Biovia, Elsevier, GSK, Novartis and Roche. The goal for UDM format is for it to become an industry-wide standard; by standardizing data, we can unlock its value as interoperable, customizable and analyzable information, and support the successful adoption of AI. Ultimately, to make data standards work, all the right people need to be in the room together—which is our goal at the Alliance.”