That’s a wrap on the 2026 GU Cancer Symposium! Excellent speakers and posters this year! #IU #Research Congrats to @Yaru00473993 on receiving the Michael O. Koch Award for Research Excellence. @asmaakenawy @ClintCaryMD

Researchers from @YaleMed & other institutions unveil imaging platform that measures metabolic activity of individual cells & identifies their specific cell type, all within a single slice of intact human tissue. Yale authors include @archie_enninful, @MinaXu7, @RongFan8 & @ZongmingMa. bit.ly/4xMAKQL

Wonderful @OpenEvidence

Memorial Sloan Kettering opening its precision oncology data to physicians nationwide through OpenEvidence. Today's Forbes has the story. @MSKCancerCenter has spent years building the world's most detailed map of which cancer mutations respond to which treatments. Until now, you had to be treated at MSK to benefit from it. Soon, an oncologist in a small practice in Montana will see the same evidence as a subspecialist in Manhattan, at the moment it matters, for the patient in front of them. More patients getting the right treatment the first time, wherever they live. Read full article here forbes.com/sites/innovationr…

Everyone's going to remember the mouse rescue from this #XunZi paper! The Chk2 inhibitor, the saved dopamine neurons, the motor deficits reversing…and fair, it's the convincing bit. But it's also the part that makes me a little uneasy about this whole genre of "AI biologist" thingy, because the thing everyone feels comfortable with is the wet lab, and the wet lab was done by people. The AI produced a ranked list. Humans decided which ten to test, ran the immunoblots, injected the mice, did the co-IP. So…what exactly are we crediting the model with, you know? Here's what I think is actually the interesting claim, and it's quieter than the headline. Two separate models that each perform mediocrely (the reasoning half around 0.65, the omics half similar) combine into something that hits 0.88, 0.92 on kinases. That's not nothing, that is real result. The "left brain/right brain" framing is (to me) marketing dressing on what is basically a sensible ensemble, but the underlying observation, that logic-over-literature and pattern-over-data catch different errors, that holds up and it's worth taking seriously. Then there's the CHK2–LRRK2 thing, which the authors lean on quite hard…a link "not in PubMed," predicted de novo, then confirmed by pulldown. Genuinely cool if true…no?. But I keep wanting to poke it. Absent from PubMed is not the same as absent from the model's training substrate, which was 24 million papers plus curated databases. A connection can be latent across a thousand documents without any single one stating it. Is that "novel discovery" or "very good retrieval of something no human had bothered to write down"? I'm not sure the distinction matters for the biology…but it matters a lot for what we claim the tool can do. And the negative-labels problem, which they're honest about, is the part nobody…well, at least not me, wants to sit with. Every uncharacterized gene gets scored as a "no”. So the model's confident negatives are really "no evidence yet”, dressed up as knowledge. In a field where the whole game is the unexplored, training on the assumption that unexplored equals irrelevant is a strange foundation, and it's exactly why the thing wobbles on rare diseases with thin literature. After reading it I’m landing somewhere unsatisfying. The paper is good, the validation is real, Chk2 looks like a legitimate PD target now. But the story we're telling…”machine reasons its way to new biology” is not quite the story the methods describe. The machine narrowed a “search space”. That's useful, maybe even transformative at scale. It's just a different, smaller claim than the one on the tin. nature.com/articles/s41551-0…

This is the closest to the mouse version of Sulston’s lineage map that anyone has gotten so far!!! Outstanding achievement!!!! Also, inspiring to see how this is the crowning achievement resulting from a decade of top tech dev!!! biorxiv.org/content/10.64898…

Congrats to Dr. Senthil Muthuswamy, new Director of @theNCI's Center for Cancer Research. His lab's work growing patient tumors as 3D organoids—testing how they respond to drugs before a patient ever starts treatment—is some of the most practical translational science we have.