[Madrid, Spain = E-Daily Reporter Kim Jinsoo ] “The technology for distinguishing cells was developed several years ago. We are now continuing our research on how to interpret that information.”

Lim Yu-ju, Medical Director at Lunit Inc.(328130), explained the future direction of artificial intelligence (AI) pathology analysis technology to Pharm E-Daily at the European Society for Medical Oncology (ESMO) conference in Madrid, Spain. The key focus is not only on developing new AI products but also on applying existing technology to various cancer types and drug development, as well as utilizing accumulated data to discover new biomarkers.

At this year’s ESMO, Lunit Inc. presented five studies utilizing its AI pathology analysis platform, “Lunit Scope,” for breast, lung, and pancreatic cancers, among others. Director Lim said, “While last year’s focus was on analyses related to immunohistochemistry (IHC) and antibody-drug conjugates (ADCs), this year there was significant interest in tumor microenvironment analysis.”

Expanding Data Interpretation by Cancer Type Through an AI Platform

Lunit Inc.’s “Scope IO” identifies tumor cells, immune cells, and endothelial cells in standard hematoxylin and eosin (H&E)-stained tissue. The technology quantifies the location and density of each cell, the distribution of the tumor-surrounding stroma, and the spatial relationships between cells.

Director Lim explained, “We trained the model to first distinguish between tumor cells and endothelial cells, regardless of the cancer type,” adding, “Once the type and coordinates of each cell are quantified, we can calculate various additional pieces of information.” Rather than developing new basic analysis technologies for each cancer type, this approach involves interpreting data obtained from the existing platform to suit specific research objectives.

The scope of data utilization is also expanding. While the initial version of Scope IO focused on immune subtype analysis—distinguishing whether immune cells had infiltrated the tumor—recent research has expanded to include the distribution of endothelial cells and the stroma, as well as the association between genetic mutations and patient survival outcomes.

A study presented at this conference involving 181 patients with pancreatic ductal adenocarcinoma is a prime example. Lunit Inc. used Scope IO to analyze the tumor-to-stroma ratio and the distribution of endothelial cells in tissue samples, then compared these findings with KRAS mutations and survival outcomes. In both the overall patient cohort and the group with the KRAS G12D mutation, the association between the tumor-to-stroma ratio and progression-free survival (PFS) showed opposite trends. This indicates that even the same histological features can have different prognostic significance depending on genetic mutations and the surrounding environment.

“There are many factors to consider regarding endothelial cells, such as their relationship with tumor or immune cells and their distance from blood vessels,” he said. “We already have the technology to detect the location of cells. We are now researching how to interpret that information.”

The prognostic significance of endothelial cell distribution was also explored in small cell lung cancer. He emphasized, “The significance of the pancreatic cancer poster lies not so much in the new technology itself, but in the ability to interpret and utilize information derived from existing technologies in various ways.”

Lunit Inc. has also developed the “Scope Genotype Predictor,” which uses morphological information from H&E-stained tissue to predict genetic mutations.

“It’s a platform that can be trained to recognize various mutations based on foundation models, rather than analyzing only specific genes,” he explained, adding, “We are also utilizing existing technology for KRAS analysis.”

Expanding into New Drug Development for Pharmaceutical Companies… Pushing for Commercialization

Image analysis technology is also being utilized in pharmaceutical companies’ new drug development. Lunit Inc. recently reached an agreement with U.S.-based Summit Therapeutics to conduct tumor microenvironment analysis research for the development of “Ibonecimab,” a PD-1/VEGF bispecific antibody. The goal is to use SCOPE IO to explore which tumor microenvironment characteristics are associated with treatment response.

Director Lim stated, “Summit wants to identify which patients are more likely to respond to Ivonesimab,” adding, “We are jointly exploring whether information on immune cells and the tumor microenvironment can be utilized as biomarkers.”

Furthermore, Lunit Inc. plans to expand its collaboration with partners from early-stage analysis in new drug development to patient selection for clinical trials and the development of companion diagnostics. For example, the company is currently collaborating with Japan’s Daiichi Sankyo to apply AI analysis technology to two anticancer drug research programs currently in development.

He explained, “The focus of our collaboration with Daiichi Sankyo is not on analyzing HER2 expression levels in existing anticancer drugs, but on identifying biomarkers for new drugs in the development stage. By applying Lunit Inc.’s AI pathology analysis platform to Daiichi Sankyo’s two new anticancer drug pipelines, our strategy is to identify biomarkers necessary for patient selection from the earliest stages of drug development.”

Finally, Director Lim stated, “The goal for all three product groups is to broaden their utility by linking biomarker discovery to treatment outcomes, which will ultimately drive revenue,” adding, “Pharmaceutical companies will continue to seek out this technology only if it can actually help patients in clinical practice.”