Pharmaceutical and healthcare companies are utilizing analytics and artificial intelligence technologies for a variety of applications such as manufacturing drugs, recommending types of patient care, and predicting patient illness.
Artificial intelligence in the pharmaceutical space is widespread, with leading US-based pharmaceutical firms like Eli Lilly and Pfizer readily deploying artificial intelligence technologies in streamlining drug development.
A Pfizer research team has recently been trying out Cortex, a collaborative, analytics and deep learning platform from Vyasa, in two pilot stages.
AI Piloting in Pharma
According to Pfizer’s director of analytics and data sciences Matt St. Louis, one pilot utilizes Cortex in categorizing particle shapes based on a drug substance’s partial microscopic images.
He also revealed that the other pilot involved feeding 15,000 images into a deep-learning model in a bid to identify the similarities that exist between them.
Vyasa, which began in Boston back in 2016, boasts deep learning technology that is designed specifically for life sciences.
Vyasa sells both Cortex and Layar, a scalable analytics data lake.
Both products allow access to the in-house deep learning and analytics tools of Vyasa.
Vyasa’s Founder and Chief Executive Officer Chris Bouton was a former Pfizer worker.
According to St. Louis, there are many existing particle shaped categories, and the team sees different outcomes based on the image quality and category.
While several shapes have success rates of nearly 90 percent, others have had a much lower rate of success.
“There are certain things that could be done on the shapes that were not very well recognized,” St. Louis said.
The research team at Pfizer leverages Vyasa in automating its workflow.
Why is Data Quality Important?
St. Louis said that data quality is among the main reasons for the low rate of success.
His boss, Pfizer’s Head of Digital Innovation and Data Science in Worldwide Research and Development (WRD) Vijay Bulusu, agrees.
“It’s a well-known fact in this field that it is less about the tools and the algorithms and more about the quality of data,” Bulusu said.
Bulusu said that St. Louis has been striving to prepare the team’s data sets, and the process has taken a long time.
“Some fundamental work has to be done on the data level,” he said.
To start with, the team was dealing with a limited volume of training data, which means that adding extra training sets could dramatically boost accuracy.
According to Bulusu, utilizing 3D images as opposed to 2D microscopic images could also improve accuracy.
He added that even though artificial intelligence in pharma is widespread, the team at Pfizer is still new to leveraging such types of analytics and deep learning tools.
Nevertheless, Bulusu said he is optimistic as far as Vyasa is concerned, and looks forward to continuing the use of Cortex, perhaps even going beyond a pilot stage.
Despite Vyasa’s potential in expediting and automating sections of the team’s workflow, Bulusu anticipates cultural obstacles.