Built to close the gap between what biomarker testing can tell you, and what patients actually need to know.

The American Cancer Society estimates 20 million new cancer cases and 9.7 million cancer deaths worldwide in 2022 (American Cancer Society, 2024). Lung cancer alone claimed roughly 1.8 million lives that year — more than breast, colorectal, and prostate cancer combined (American Cancer Society, 2024).

Lung cancer may be treated with surgery, radiotherapy, chemotherapy, targeted therapy, immunotherapy, or some combination of these — but treatment options and their effectiveness vary according to the patient's cancer type, stage, and its molecular, biological, and physiological characteristics.

Non-small cell lung cancer in particular has historically seen poor treatment and survival outcomes, largely because it's so often diagnosed at a late stage (Siegel, 2023).

Immunotherapy has shown real promise over the last decade, improving outcomes across many cancer types. Checkpoint inhibitors that target PD-L1 block some of the mechanisms cancer cells use to evade the immune system — but that same mechanism can sometimes suppress immune responses the body needs for other things, causing serious side effects.

The catch: doctors currently don't have a reliable way to know in advance which patients will respond well to a given immunotherapy, and which might instead experience serious immune-related adverse events. In an analysis of 45 FDA immunotherapy approvals, the standard biomarker used to gauge immunotherapy response and toxicity — PD-L1 — was predictive in only 28.9% of cases, and not predictive in a further 53.3% (Davis & Patel, 2019). Relying on a single biomarker has real limits, both for treatment success and for anticipating the adverse reactions a patient may face from a therapy that doesn't suit their cancer.

That gap is why OnceraPlus exists.

We built OnceraPlus to help close that gap: a model trained specifically to stratify pre-treatment risk using the clinical factors already available in a patient's workup, so the decision is better-informed from the start.

OnceraPlus estimates pre-treatment immunotherapy toxicity risk using clinical factors that are typically already part of a standard workup — no additional biopsy, no new tests to order, nothing extra for the patient to go through.

It wasn't adapted from a general-purpose AI system; it was built for this one task from the start, trained and validated on clinical outcomes from NHS partner hospitals, and it's MHRA registered.

We're starting with lung cancer because that's where the need and the data lined up first. We're expanding both — more hospital partnerships, more disease models — as the work continues.

Built for this, from the ground up. OnceraPlus isn't a general-purpose AI system adapted for healthcare. It was built specifically to predict immunotherapy toxicity risk, trained and validated on clinical outcomes — a meaningful difference from asking a general AI assistant a health question.

Trained on NHS patient data. The model is built on clinical data from more than 3,999 patients, including 500+ with lung cancer who received immunotherapy.

MHRA registered.

Working with NHS hospitals; partnerships in progress. We're currently working with a number of NHS hospitals as we expand the data behind our models.

Research evidence. Our approach to predicting immunotherapy response and toxicity has been presented at leading oncology conferences and published as conference abstracts:

• Journal of Clinical Oncology (2023 ASCO Annual Meeting) — a machine learning algorithm to predict immunotherapy response in small cell and non-small cell lung cancer
• Annals of Oncology (ESMO 2023) — companion abstract on the same lung cancer response model
• Journal of Clinical Oncology (2024 ASCO Annual Meeting) — immunotherapy toxicity prediction in melanoma using machine learning
• Clinical Cancer Research (AACR 2025) — a precision-medicine approach to melanoma immunotherapy, predicting response, adverse events, and hospital admissions using explainable AI