By Allison Proffitt
December 3, 2020 | In a partnership seeking to exploit the genotype-phenotype connection, Agendia and Paige have announced a strategic partnership to revolutionize treatment planning for breast cancer using AI and pathology images.
Paige is an AI company founded in 2017 by scientists at Memorial Sloan Kettering Cancer Center and spun out in 2018. The company started digitizing the Memorial Sloan Kettering pathology slide archive. They’ve digitized over three million slides so far and continue to add about 100,000 each month. Digitizing, or photographing, slides makes all the image data available to computer vision—and computer vision is good at pattern recognition. The company founders saw an opportunity to train an artificial intelligence platform to spot cancer using the vast quantity of pathology slides in the archive.
Instead of extensively annotating the slide archive, Paige used only slides paired with reported diagnoses for training. They published their deep learning system in Nature Medicine in July 2019 (DOI: 10.1038/s41591-019-0508-1). “We developed a novel framework that leverages multiple instance learning to train deep neural networks, resulting in a semantically rich tile-level feature representation,” the authors wrote. “These representations are then used in a recurrent neural network (RNN) to integrate the information across the whole slide and report the final classification result.” In the evaluations of 44,732 whole slide images from 15,187 patients from 800 hospitals, the Paige technology allowed pathologists to exclude 65–75% of slides while retaining 100% sensitivity, the team reported.
“Paige’s technology will compare all of the patterns in that tumor and the microenvironment of the tumor to a database and be able to show the pathologist: is there a higher likelihood of cancer at this location or at that location,” Leo Grady, the company’s CEO told Diagnostics World.
The platform has already received regulatory stamps of approval. In July 2020, Paige received Federal Drug Administration (FDA) 510(k) clearance for the FullFocus, a digital pathology image viewer for the purpose of primary diagnosis. In November 2019, Paige received CE Mark for Paige Prostate, its first-in-class prostate cancer detection solution, and Paige Insight, its AI-native digital pathology viewer, both for primary diagnosis. It has seen funding support as well, with $95M in total capital raised, including $15M in July 2020 to close its Series B round.
But Paige believes a different kind of training dataset will tease out even more actionable health information from slides. That’s where Agendia comes in.
Agendia was founded in 2003 as a spin-off of the Netherlands Cancer Institute in Amsterdam and is now headquartered in Irvine, Calif. Its flagship genetic test, MammaPrint, was launched in Europe in 2004 and in the U.S. in 2007. MammaPrint uses a 70-gene panel to predict breast cancer recurrence. BluePrint, a laboratory-developed test performed by Agenda, is an 80-gene molecular subtyping assay to identify breast tumor growth drivers. Together, the two tests are used to develop treatment plans for individual breast cancer patients.
The idea is that the pathogenic mutations Agendia has been testing for are visible in the tissue. Those genotypes are expressed phenotypically. “This has been shown to be true in many cases with certain molecular changes at the tissue level,” Grady said. “For years pathologists have been saying, ‘I can see an FGFR mutation. I can see HER2 positivity in this tissue.’ But they haven’t been able to quantify it. They haven’t been able to standardize it… The kind of AI technology that Paige has built—and the databases that we’ve [developed] with MSK and Agendia—have allowed us really to find, qualify, and standardize those patterns that pathologists have been seeing for years.”
In last month’s partnership announcement with Agendia, Paige is adding Agendia’s database of test results to the AI platform. “The partnership with Agendia we really see as the first of the commercializations of this technology,” Grady explained. “We’re looking at patterns in the tumor and comparing that to Agendia’s database.”
The new partnership will give Paige access to all the MammaPrint and BluePrint assays, letting the Paige platform incorporate the genetic data that Agendia has been using to make recurrence predictions and identify tumor growth drivers.
“The AI will catalog and learn from each tissue sample that is brought in for it to evaluate, as well as providing faster results and helping to identify biomarkers that will help guide early treatment planning,” explained Mark Straley, CEO of Agendia. He expects it to be “transformational.”
The AI will be incorporated into the MammaPrint and BluePrint tests, Straley explained, not “add-ons” that physicians will have to choose. The new AI-enabled MammaPrint and BluePrint test results will be returned on the same day they were ordered most cases, he said, must faster than the current five-day. Treatment plans can be finalized and initiated much faster. The digitized tests will also expand testing access, “in countries where tissue ‘send out’ is not allowed,” he added.
“Initially, the value propositions around the partnership will be tissue retention, same day turnaround for test results (in most cases), opportunity for revenue generation at partner labs and access to genomic testing in markets where tissue send out is not allowed,” Straley said. “In the future, the integration of AI and genomic results may provide new insights that allow us to better stratify patients for treatment planning.”
“The knowledge base will continue to expand with each and every MammaPrint or BluePrint assay that is evaluated with the Paige Platform,” he added. “Our goal is to move beyond early intervention and develop innovative new tests to support treatment planning further along the care continuum—disease monitoring following definitive treatment and metastatic treatment planning. We are committed to being a partner for our physicians and their patients throughout the entire patient journey.”
Paige expects that partnership to deliver new understandings too and foresees the partnership with Agendia as the first in a panel of AI-enabled tests.
“There’s an incredible amount of information contained in this tissue. So being able to extract that information out of AI and match that not only to diagnoses but also to what treatments are going to work for that patient is really a new field,” Grady said.