NEW YORK – A new partnership between Roche Tissue Diagnostics and digital pathology firm PathAI aims to develop artificial intelligence-based algorithms and companion diagnostic tests, tying into both companies' larger digital pathology strategies.
Roche is "very committed to the belief" that anatomic pathology is moving to a fully digitized workflow, said Michael Rivers, the lifecycle leader for digital pathology at Roche. In recent years, the company has seen investments in AI solutions for digital pathology grow rapidly, as has interest in the space, "and that interest has really been growing on the pharma side," he added.
While most of that has been in early pharma research, there has also been increasing interest in developing companion diagnostic algorithms combined with assays.
The partnership with Boston-based PathAI, which was announced earlier this month, focuses on companion diagnostic AI solutions for assays that Roche Tissue Diagnostics is developing. Those AI solutions are "very complex" and take "a significant amount of expertise and resources," so collaborating with PathAI to develop those algorithms is a "great way to go after this opportunity," he said.
Eric Walk, PathAI's chief medical officer, said that Roche wanted "a seamless way to incorporate AI into the already established companion diagnostics development process," and that the partnership will allow both companies to work with pharmaceutical partners together and explore early phases of drug and companion diagnostic development. He added that the companies will be able to comarket an integrated solution to pharma partners, meeting with them together to identify appropriate projects and jointly developing complete solutions.
He noted that most tissue-based companion diagnostics are based on immunohistochemistry to visualize proteins, which are manually interpreted and scored. The "vision for the future," however, is to create predictive biomarkers and scoring systems that are "even more precise" to select responders to certain treatments.
The goal of the partnership, PathAI's Walk said, is "to codevelop the assay component of the CDx together with the AI algorithm component" instead of "simply adding the algorithm at the end of assay development." As an example, he cited the biomarker scoring system, "which today is chosen in part based on early clinical efficacy data but also in part based on what human pathologists can interpret reproducibly." AI-based methods, meantime, "have the precision to reproducibly measure the optimal biomarker cutoff point, even if that is an odd number" that wouldn't be feasible via manual methods.
AI tools can be used to "better measure established biomarkers with more accuracy, reproducibility, and standardization, as well as discover new spatial relationships that may be difficult or impossible to recognize by the human eye alone," PathAI CEO Andy Beck added. Walk said that the partnership will focus first on IHC assays but may also include biomarker discovery, as well.
PathAI's role in the partnership will be focused on the development of the AI algorithms, with Roche focused on the development of the companion diagnostic assay, but the development processes "will be tightly coordinated in parallel to generate the best overall CDx solution," Walk said. PathAI typically develops its algorithms to "be robust across multiple image formats," he added.
Walk noted that AI could also influence how the assays are developed, because "right now, assays are developed with human eyes in mind," meaning they must be interpretable by humans. That restriction goes away with the use of AI, and the lack of restrictions and increased precision afforded by artificial intelligence tools could shorten the development time of the CDx solutions and provide more freedom to explore the different dynamic ranges of each assay.
Roche has a "well-established" companion diagnostic business and has seen interest from external pharmaceutical partners in AI solutions not only for discovery but also as products combined with CDx assays to "assist the pathologist in the interpretation of the assays and ultimately to guide the therapy," Rivers said.
Under the agreement, Roche will retain leadership of the overall companion diagnostic program, developing its own CDx assays for pharma partners and then using either internal resources or PathAI to develop image analysis software that will ultimately be a single product including the AI solution and the CDx assay, Rivers said.
All of the products developed under the agreement will be deployed on Roche's Navify image management solution. The Navify platform is a software environment that pathologists can use to manage their workflow, including getting access to image analysis tools such as AI-based algorithms to help see areas of interest on a digitized slide or provide quantitation. The platform includes Roche-developed products as well as external third-party algorithms. PathAI was the first partner on the Navify open environment, Rivers noted, and Ibex Medical Analytics also has algorithms available on the platform. Rivers added that Roche is working with other undisclosed partners to onboard algorithms onto Navify, as well.
While the agreement is focused on companion diagnostics, the companies have discussed additional algorithm content from PathAI that could be deployed on Navify and integrated into laboratories sooner — companion diagnostics are long-term projects, but there is an "opportunity to bring AI content onto our platform sooner," he said.
The CDx algorithm agreement is also exclusive to PathAI, Rivers noted. During the course of the agreement, the terms of which are undisclosed, Roche will only work with PathAI or its internal teams to develop CDx algorithms, and PathAI will only work with Roche on companion diagnostic assays.
Because the algorithms and assays are submitted as one device under one label, it "requires intense and very close cooperation and collaboration, and we believe that we will be able to more efficiently serve our pharma partners and, ultimately, patients if we're able to focus on a single, very high-quality partner like PathAI," Rivers said. Roche will handle the regulatory and commercialization aspects of any products that come out of the partnership, he added.
PathAI's Beck noted that a "central pillar" of PathAI's strategy has been building capabilities to support companion diagnostic work, and this partnership ties into that goal. Drug development and diagnostics have been key components of the firm's work since its founding, and the firm received 510(k) clearance from the US Food and Drug Administration in 2022 for its AISight Dx digital pathology platform for primary diagnosis in clinical settings.
"As precision medicine continues to advance, we're going to want even better technologies to identify subsets of patients," he said. For tissue-based companion diagnostics, "in the future, AI will just continue to play an increasingly important role."
While digital pathology continues to see growing interest among laboratories, barriers still remain to its adoption. In Rivers' view, the partnership "will accelerate our ability to penetrate" the digital pathology market and accelerate the digital transformation of pathology labs.
Navify has been in a significant investment phase, but after transitioning to the cloud and the launch of a high-volume digital slide scanner, Roche has recently seen a big uptick in sales of the solution, and Roche expects that will continue to accelerate. "I see AI as a really important longer-term driver in digitization, and I think this collaboration will help us unlock the potential there for these CDx solutions," Rivers added.