Altis Labs Raises $25 Million Series A

Altis Labs of Toronto, Canada has raised a $25 million Series A. According to the company's September 17 announcement, OrbiMed and Qiming Venture Partners USA co-led the round, with participation from Innovation Endeavors, Benchstrength, Fusion Fund and Lumira Ventures' Cancer Breakthrough Fund among others. The company said the money will go toward extending its models to additional cancer types, scaling commercial deployment with global pharmaceutical partners, and establishing AI oncology trial endpoints as a new standard.

What it sells is not a drug but a measure. Altis's flagship model, IPRO, predicts a patient's survival using only the CT images a trial is already acquiring. Unlike the conventional approach of measuring the size of target lesions, it reads the whole three-dimensional scan and quantifies thousands of prognostic imaging biomarkers spanning tumour burden, body composition and organ health. The training data is a multi-centre database combining longitudinal imaging, molecular, clinical and outcomes data, covering more than 500,000 patient years, the company says.

Why such a measure exists comes from an old problem in trials. Oncology studies have long leaned on imaging endpoints such as objective response rate, yet that number does not fully say whether patients end up living longer. Developers still have to use an interim measure to decide whether to continue or stop, how to design the next trial, and which programme to fund. Each of those calls carries hundreds of millions of dollars in research and development spending and billions in commercial value, the company wrote.

The place where the numbers were checked is the MARIPOSA Phase 3 trial. In a post-hoc analysis run by the trial sponsor itself, IPRO automatically read roughly 10,000 CT images from previously untreated patients with EGFR-mutated advanced non-small cell lung cancer. IPRO Response was defined as a 50% or greater improvement in the IPRO-alpha score from baseline, and from week 16 the investigational arm ran ahead of the control arm. The response rate ratio was 1.33, with a 95% confidence interval of 1.03 to 1.71. In the same trial, objective response rate never anticipated the median overall survival benefit of more than 12 months.

The gap in time is the company's entire claim. According to the announcement, IPRO Response Rate picked up a statistically significant treatment effect 11 months before the primary progression-free survival readout and 26 months before the final overall survival readout. Pooling patient-level trajectories over time, patients whose IPRO deteriorated did worse on survival and those whose IPRO improved did better. The results were presented on September 13 at the World Conference on Lung Cancer 2026 in Seoul as poster P1.046.

Felix Baldauf-Lenschen, Founder and Chief Executive Officer of Altis Labs, said: "Every clinical trial is fundamentally a learning process, and right now it is a slow one. The faster and more accurately we can determine whether a treatment is truly delivering benefit to patients, the faster our biopharma partners can bring new effective medicines to those who need them." Tal Zaks, Partner at OrbiMed, said: "Having spent the majority of my career developing cancer therapies, I know how much rides on early efficacy signals that can often fail to predict whether a drug will truly improve patient outcomes."

What the investors bought is data more than a model, and the announcement shows it. Zaks said Altis had assembled the industry's most comprehensive longitudinal oncology dataset linking imaging to outcomes, and had translated it into a novel outcome measure with Phase 3 validation behind it. Anna French, Managing Partner at Qiming Venture Partners USA, said measuring efficacy earlier and more accurately is the most impactful application of AI in drug development. High late-stage failure rates, costs and timelines are the bottleneck in therapeutic research and development, and AI oncology endpoints aim straight at that spot, she said.

The numbers behind the disease explain the scale too. Non-small cell lung cancer accounts for up to 85% of lung cancer cases worldwide and is the leading cause of cancer death. EGFR alterations are among the most common driver mutations, found in 10% to 15% of Western patients and up to 40% to 50% of Asian patients with adenocarcinoma histology. Despite advances in targeted therapy, the five-year survival rate for advanced EGFR-mutated non-small cell lung cancer remains under 20% on standard of care.

In the company announcement METAL read in full, IPRO is described as being used by top 20 pharmaceutical companies to analyse data from phase 1 through phase 4 trials. According to reports, the round was relayed again on September 21 through an industry outlet, at the same $25 million. METAL has reported that Alibaba released RADAR, an abdominal CT diagnostic model, and has reported that an AI-designed drug showed an effect across six ageing clocks. The side that reads images and the side that makes drugs are meeting in the same data.

Seen through an engineer's eyes, the lines this company still has to cross are clear. A post-hoc analysis is run with the results already known, so whether the same finding holds in a prospective design is for the next trial to say. The 50% improvement threshold that defines IPRO Response is one choice, and a value that has to be set again for each cancer type. Three things are worth watching next. Whether regulators accept this as a formal endpoint, whether the same time gap reproduces outside lung cancer, and whether use by top 20 pharmaceutical companies moves beyond analytical support into the room where continue-or-stop decisions are made.