The Algorithm That Went Shopping for a Cancer Drug

But those drugs come with baggage. The biggest issue is peripheral edema, a painful swelling in the arms and legs that affects roughly 62% to 82% of patients on existing MET inhibitors. It's not just uncomfortable; it's the kind of side effect that makes patients want to quit treatment.

DO-2's early data tells a very different story. In its Phase 1 trial of 29 patients, the peripheral edema rate was just 5%. One patient. Grade 1 (the mildest category). Zero Grade 4 adverse events across the board.

On the efficacy side, every single evaluable MET exon 14 skipping NSCLC patient who received a therapeutic dose saw their tumor shrink. That's 10 out of 10. Now, this is early-stage data from a small trial, so treat it like a promising first date rather than a marriage proposal. But the combination of strong activity and a dramatically cleaner safety profile is exactly what gets oncologists (and algorithms) excited.

The Deuterium Trick

DO-2 has an unusual secret weapon baked into its chemistry: deuterium.

Deuterium is just a heavier version of hydrogen. Same element, extra neutron. When you swap specific hydrogen atoms in a drug molecule for deuterium atoms, you get a bond that's harder for the body's enzymes to break. It's like replacing a paper lock with a metal one; the key still fits, but it takes longer to turn.

This matters because it changes how the body processes the drug. DeuterOncology designed DO-2 with "fast on, fast off" binding to MET. The drug delivers about 8 to 12 hours of potent MET inhibition per day, enough to kill cancer cells but not long enough to cause the chronic endothelial damage thought to drive all that swelling with older drugs.

The original scaffold traces back to Janssen Pharmaceutica, later licensed to OCTIMET Oncology, and then DeuterOncology grabbed the worldwide rights to the deuterated version. The strategy of applying deuterium to validated drug scaffolds already has real precedents: donafenib (a deuterated version of sorafenib) is approved in China for liver cancer, and deucravacitinib proved the concept outside oncology.

DO-2 also appears to be brain-penetrant in preclinical work. That's a big deal for lung cancer patients, since brain metastases are common and notoriously hard to treat. On top of that, DO-2's active metabolite reportedly hits a second target in the RAS signaling pathway, which could help delay resistance to MET inhibitors. Two mechanisms from one pill.

A Crowded Dance Floor

DO-2 isn't walking into an empty room. Capmatinib and tepotinib own the Western markets. Savolitinib dominates China, especially in combination with osimertinib for patients whose EGFR-targeted therapy stopped working because of MET amplification.

Then there's AbbVie, which got the first MET-targeted antibody-drug conjugate (ADC) approved in 2025: telisotuzumab vedotin. ADCs work differently; they deliver toxic payloads directly to cancer cells expressing the MET protein, expanding the eligible patient pool beyond just those with specific genetic mutations.

A wave of Chinese competitors (glumetinib, vebreltinib, dalmelitinib) adds even more pressure. DO-2's path to standing out depends on confirming that safety advantage in larger trials and proving its resistance-busting metabolite does what preclinical models suggest.

Patent protection through December 2040 gives Pathos a decent runway. But the dose expansion cohort, now enrolling first-line MET exon 14 skipping NSCLC patients at 60 mg once daily across sites in Belgium, the Netherlands, and France, will be the real test.

The Bigger Bet

This deal is really two stories in one.

The first is about DO-2 itself: a clever piece of chemistry that might genuinely be better than what's available for MET-driven lung cancer. The early data is striking, but "10 out of 10 patients" in a Phase 1 trial is a trailer, not the movie.

The second story is about how deals get done. Pathos AI raised $365 million in its May 2025 Series D at a reported $1.6 billion valuation. The company has roughly $467 million in total funding, a CEO (Iker Huerga) who previously led oncology data science at AstraZeneca, and a thesis that AI can systematically find undervalued clinical-stage drugs better than humans with Rolodexes.

If Foundry keeps surfacing winners, the "search and buy" model could genuinely change how biotech M&A works. Pharma BD teams already use AI for scouting, but Pathos is building the entire company around letting the algorithm drive. It's the difference between using GPS for directions and letting a self-driving car take the wheel.

The risk? Algorithms are only as good as their data and assumptions. A 95% failure rate for enterprise AI pilots (per a widely cited MIT study) should keep everyone appropriately humble. But with clinical-stage assets in its portfolio, including P-300 for prostate cancer and P-500 for brain tumors, Pathos is putting real money behind the thesis that machines can pick drugs better than people.

Wall Street (and the oncology community) will be watching the DO-2 expansion data closely. If those edema rates hold and tumors keep shrinking, the algorithm might deserve a promotion.

The $2 Stock That Just Sold for $1.1 Billion

Esperion Therapeutics once traded above $115. It was languishing under $2 when ARCHIMED swooped in with a $1.1 billion take-private deal. The cholesterol drug everybody forgot about might be the smartest private equity bet of the year.