Caris-style AI biomarker work is advancing in Non-Small Cell Lung Cancer (NSCLC): Onc.AI said its CT-based Onclara IO model stratified 12-month survival in 205 patients with PD-L1–high, mutation-negative metastatic NSCLC, with median overall survival of 484 days in the high group vs. 155 days in the low group. The external blinded validation study will be presented at ASCO, and the company says the tool can return a risk report from a baseline CT scan within 24 hours.

Why It Matters To Oncology

• Onclara IO is positioned as a noninvasive imaging biomarker that could complement PD-L1 testing when selecting first-line immunotherapy.

• For clinicians, the appeal is rapid risk stratification from routine pretreatment CT imaging, potentially helping identify which PD-L1–high patients are most likely to benefit.

• For drug discovery and development, imaging-based biomarkers may help enrich trials, refine prognostic subgroups and support biomarker-informed treatment strategies without requiring additional tissue.

The Financials

• Onc.AI said it plans to submit Onclara IO for software-as-a-medical-device clearance next year.

• The company is also building on prior metastatic NSCLC work, including its FDA breakthrough-designated Serial CT Response Score model.

• That broader platform story could matter commercially if imaging AI tools show utility across treatment selection, response assessment and survival prediction.

What They're Saying

• Study author Ravi Parikh said the findings position Onclara IO as a noninvasive imaging biomarker that complements PD-L1 testing and could inform first-line treatment selection and risk-adapted strategies.

• Flatiron Health's Jacqueline Law said the results reinforce the value of combining curated real-world data with imaging AI to support more precise, biomarker-informed decisions.

What's Next

• The immediate milestone is the ASCO presentation, where clinicians will look for more detail on model performance, subgroup behavior and potential confounders in the retrospective cohort.

• Regulatory submission is planned for next year.

• Key questions ahead: prospective validation, generalizability across scanners and practice settings, and whether the biomarker changes treatment decisions or outcomes in real-world use.