At the European Society for Medical Oncology conference (ESMO Congress 2025), Lunit presented new clinical evidence for its Artificial Intelligence (AI) pathology analysis.

The results are particularly significant as Lunit demonstrated that AI can distinguish treatment response in a subset of proficient mismatch repair (pMMR) colorectal cancer (CRC) patients, a group that typically shows little response to immunotherapy.

DailyPharm met with Im Yu-ju, Medical Director of Lunit's Oncology Group (Hematology-Oncology Specialist), at ESMO 2025 to discuss the clinical significance and future vision of the research.


"Predicting treatment response from a single Slide... AI biomarker proves utility"

Lunit presented two abstracts utilizing its AI biomarker platform, Lunit SCOPE.

In particular, Oral Presentation detailing the results of a joint study with Professor Chiara Cremolini's research team at the University of Pisa garnered attention.

The study's core objective was to predict the therapeutic effect of atezolizumab (Tecentriq) combination therapy using Lunit's AI pathology platform, 'Lunit SCOPE.'

Im explained, "pMMR colorectal cancer is a notoriously intractable cancer that barely responds to immunotherapy. However, this study allowed us to identify a specific patient subgroup that benefits from the addition of immunotherapy."

Im emphasized, "We quantified the tumor microenvironment (TME) using only conventional H&E (Hematoxylin & Eosin-stained slide) slides, without the need for new tests or tissue collection, to predict treatment response. This result shows that AI can possess clinical utility as a biomarker."

H&E slides are standard stained tissue slides made from most patient specimens during pathological diagnosis. The clinical applicability is high because they can be used without additional testing.

In the study, researchers analyzed pathology slides from 161 patients using Lunit SCOPE to quantify the density of six cell types, including lymphocytes and tumor cells. Subsequently, they stratified patients into two groups (A/B).

The analysis showed that in the atezolizumab combination group, Group A demonstrated improvement in both progression-free survival (PFS) and overall survival (OS) compared to Group B. Notably, this difference was observed only in the combination arm, but not in the chemotherapy monotherapy group, proving Lunit SCOPE to be an immunotherapy-specific predictive indicator.

"AI incorporated complex immune response...AI refines tumor microenvironment analysis"

The key finding of this research is its complex interpretation of the tumor microenvironment (TME), moving beyond simple cell density analysis.

Im stated, "The response to immunotherapy is a comprehensive result of complex immune factors, including T-cell infiltration, antigen presentation pathways, and neoantigens, not just PD-L1 expression." She added, "Lunit SCOPE quantifies these multi-layered variables through AI pathology analysis and presents them in an interpretable format."

In particular, this model moved beyond the traditional inflamed-centric classification by incorporating the interplay between various cells, including endothelial cells and fibroblasts.

Im added, "Along with lymphocyte distribution, the proportion of dividing tumor cells was the highest contributor to predicting response," and said, "AI has overcome the limitations of conventional single-factor-based biomarkers."

Furthermore, Im said, "Although it varies by slide size, the analysis of a single slide typically takes 5–10 minutes. Even large-volume data can be processed within tens of minutes." She stated, "This speed allows the results to be immediately referenced concurrently with the clinical interpretation process."

Lunit, which has attended ESMO for five consecutive years since 2021, is leveraging this research to expand collaboration discussions with global pharmaceutical companies.

Im said, "Since many immune checkpoint inhibitors are already approved, collaboration is more active in companion diagnostics for subsequent indications and new drug development, rather than new clinical trials," and added, "We are accumulating evidence through investigator-initiated clinical trial data,"

"We are concurrently researching biomarkers for next-generation anti-cancer drugs like BiTE (Bispecific T-cell Engager) and ADC based on Lunit SCOPE IO," said Im and mentioned, "Our goal is to establish the technology as a practical treatment predictive tool through collaboration with partner pharmaceutical companies from the clinical trial stage."

"AI, a New Partner in Drug Development... Expanding to ADC and TKI"

At the annual ESMO conference, presentations on new modalities, such as ADCs, are consistently featured alongside those on immunotherapies. Lunit is also expanding its AI biomarker research in these areas.

Im stated, "In the ADC field, global pharmaceutical companies are actively adopting digital pathology AI-based Companion Diagnostics (CDx), where Lunit's analysis technology can significantly contribute."

Lunit is currently developing biomarkers for ADC drugs using IHC analysis of immunostained slides and is also building a TKI response-prediction model using morphological pattern analysis.

Im also added, "If AI quantifies drug-specific response patterns, it will allow us to connect both companion diagnostics and patient-specific treatment in the future."

At this ESMO Congress 2025, Lunit also presented research on renal cell carcinoma and non-small cell lung cancer in addition to colorectal cancer.

In the renal cell carcinoma study, the immune-activated patient group showed a significantly higher ORR of 60.5% in the nivolumab + ipilimumab combination therapy compared to the non-activated group (23.2%). In the NSCLC study, the immune-activated phenotype showed a superior response in a Japanese multi-center patient cohort, confirming the reproducibility of the AI model.

Im said, "AI pathology analysis is not limited to a specific cancer type." She added, "We are conducting multi-cancer expansion studies to apply it to early treatment stages and adjuvant therapies."

Ultimately, the assessment is that AI is no longer a future technology but becoming established as a practical tool that is changing treatment strategies in real-world clinical settings.

Im concluded, "AI pathology analysis is not limited to a specific cancer type. Lunit is conducting multi-cancer expansion studies so that it can be applied to early treatment stages and adjuvant therapies."