• Noetik's TARIO-2 platform analyzed routine pretreatment H&E images from 113 BOT+BAL-treated patients and identified spatial tumor microenvironment patterns associated with response and survival
  • In a retrospective analysis of refractory MSS metastatic colorectal cancer without active liver metastases, the AI-identified subgroup had a 64% response rate to BOT+BAL, compared with 9% in the remaining cohort
  • Overall survival was significantly improved in the AI-identified MSS mCRC subgroup, with median overall survival not reached and a hazard ratio of 0.18 versus the remaining cohort

Agenus Inc. (NASDAQ:AGEN), a leader in immuno-oncology innovation, today announced new retrospective data showing that Noetik's artificial intelligence-based TARIO-2 model identified spatial tumor microenvironment patterns associated with clinical outcomes from routine pretreatment tumor pathology images in patients treated with botensilimab (BOT) plus balstilimab (BAL), Agenus' investigational next-generation multifunctional, Fc-enhanced anti-CTLA-4 and anti-PD-1 immunotherapy combination.

The data will be presented on May 30, 2026, by Ryan Dalton, Ph.D., of Noetik, during a poster session at the 2026 American Society of Clinical Oncology (ASCO) Annual Meeting. The presentation, titled "Artificial intelligence foundation model as a predictor of efficacy of next-generation checkpoint inhibition with botensilimab (BOT) + balstilimab (BAL) in solid tumors using pretreatment H&E images," evaluated whether Noetik's TARIO-2 model could analyze standard hematoxylin and eosin (H&E) pathology images to identify spatial tumor microenvironment patterns associated with clinical outcomes following treatment with BOT+BAL.

BOT is an Fc-enhanced anti-CTLA-4 antibody designed to broaden anti-tumor immune activity through effects on T-cell priming, antigen presentation and regulatory T cells within the tumor microenvironment. Given BOT+BAL's differentiated mechanism and prior observations that clinical activity is not strongly associated with traditional biomarkers such as PD-L1 expression or tumor mutational burden, broader tumor microenvironment-based approaches may be important for identifying patients most likely to benefit.

The analysis included 113 efficacy-evaluable patients treated with BOT+BAL in the C-800-01 Phase 1b trial who had available pretreatment H&E images. Tumor cohorts included microsatellite stable (MSS) metastatic colorectal cancer (mCRC) without active liver metastases, ovarian cancer and sarcomas.The analysis evaluated TARIO-2's ability to predict clinical endpoints including best overall response and overall survival.

In the MSS mCRC without active liver metastases cohort, TARIO-2 demonstrated statistically significant predictive performance for both best overall response and overall survival. Supportive trends were observed in the ovarian cancer and sarcoma cohorts. In the MSS mCRC without active liver metastases cohort, TARIO-2 also outperformed benchmark pathology foundation models in predicting best overall response and overall survival.

TARIO-2 does not rely on a traditional single-marker biomarker approach. Instead, the model applies AI-based spatial tumor microenvironment analysis to standard H&E pathology images, which are routinely generated during cancer diagnosis and clinical evaluation. By using widely available H&E images, TARIO-2 is designed to extract biologically relevant tumor microenvironment features without requiring more complex tissue-profiling approaches that may be difficult to implement routinely. This approach may support future patient stratification strategies if prospectively validated.