Travel applications for our upcoming workshop, ‘Cell State Transitions and Fate (in)decisions,’ are due in one week!
Don’t miss your opportunity to join a community formalizing a unified framework for understanding plasticity and cell state-fate relationships
www.nitmb.org/cell-state-t...
Just in: Supreme Court rules against law banning conversion therapy for minors.
8-1 decision.
Justice Ketanji Brown Jackson was the lone dissenter. www.huffpost.com/entry/suprem...
Breaking news: A federal judge found that the IRS violated federal law “approximately 42,695 times” when it shared confidential taxpayer addresses with immigration enforcement officials last summer.
Shame on NYU Langone for ending gender-affirming care & kicking trans New Yorkers to the curb. It's a violation of their promise to treat all people equally, kindly, and respectfully — and a violation of NYC's human rights law as well. Bending the knee is bad; denying people care is even worse.
The nazis redistributed the stolen property of deported Jews to the German population. It was called Aryanization.
This is no different.
en.wikipedia.org/wiki/Aryaniz...
Out in @physrevlett.bsky.social: @sfuphysics.bsky.social postdoc Antonio Patrón Castro (in tight collaboration with my colleague John Bechhoefer) pushes our information engine to higher dimensions!
doi.org/10.1103/xwjd...
So excited to be co-organizing this!! If you're in the Chicago area, I highly recommend you join us for this 2.5 day workshop -- we have stellar speakers talking about all sorts of theory and biology. Make sure to apply before the deadline if you are interested!
Excited to share new work by Daiki and Hector in collaboration with Monika: arxiv.org/abs/2512.13859 We introduce a gating mechanism in classic associative memory models and find capacity is increased far above the Hopfield limit without the usual catastrophic breakdown.
Super happy to be associated with this work !!
NEW PREPRINT led by stellar postdoc
@biophysben.bsky.social We turn diffusion models into inference machines. Score function→exact likelihoods, uncertainty estimate & interpretablity. Diffusion models generate. But classify? Untapped—until now @chanzuckerberg.bsky.social shorturl.at/s32uX
Big thanks to @goyallab.bsky.social, @vaikuntsuri.bsky.social, and Ekta Prashnani, as well as @czbiohub.bsky.social, @nitmb.bsky.social, @nufeinbergmed.bsky.social, @nucsb.bsky.social
Why?
Diffusion models are great generators—but their potential for classification & regression has barely been tapped.
And current deep classifiers?
• Hard to interpret
• Poor at uncertainty
• Don’t understand distribution geometry
We help address all three.
Today we’re excited to share Keeping SCORE, a framework that turns any diffusion model into a thermodynamic inference machine.
Using only a model’s score function, we compute exact class likelihoods and uncertainty—and even decompose which features drove the decision. www.biorxiv.org/content/10.1...
Why?
Diffusion models are great generators—but their potential for classification & regression has barely been tapped.
And current deep classifiers?
• Hard to interpret
• Poor at uncertainty
• Don’t understand distribution geometry
We help fix all three.
Not sure why @lpachter.bsky.social did not post this here. But it is brilliant. Single cell genomics finally makes it to the clinic.
excited to announce that this January we are relocating to Yale to start a theory initiative at the Center for Systems and Engineering Immunology (CSEI): @yalecsei.bsky.social
looking to hire 2 postdocs to join us in this endeavor.
more details can be found here: sites.google.com/uw.edu/statp...
Babies who are threat to DHS = 0
DHS teargas grenades thrown at babies = More than 0
(Before anyone blames adults for bringing baby there, video shows woman pull baby from her obstructed car to escape area. No indication she was there for a protest.)
www.nytimes.com/2025/09/29/u...
Trying to log into Eduroam
There is no world in which masked and armed agents of the state running down delivery drivers makes us safer. That our information environment and political leaders have convinced people otherwise is a massive failure.
I finished my estimate on required compute to make an atomic-resolution virtual cell: 10^38 FLOPs to simulate a human cell for 1 day. We should be able to do this simulation in 2074 using 200 TW of power. 1/3
Researchers usually target genes that are highly over or under expressed in variant (disease) cells relative to baseline. We go beyond such differential expression analysis to probabilistically match perturbation-induced transcriptional shifts to measured changes from healthy and diseased tissue.
Excited to share my new work on using AI and measured transcriptional responses to gene perturbations (knockdowns for example) to predict which sets of genes cause/mitigate complex disease phenotypes like cancer metastasis, IBD, or asthma, out now in PNAS! www.pnas.org/doi/full/10....
A whistleblower complaint says that the personal data of over 300 million Americans was copied to a private cloud account to allow access by members of the Department of Government Efficiency team.
This is not just HHMI. All Chan Zuckerberg Initiative funding calls are CLOSED. The Gates Foundation has no Grand Challenges funding opportunities. I've never seen anything like this. Private foundations have just decided to stop supporting biomedical research.
A reminder that the mRNA vaccine technology saved millions of lives during the Covid pandemic and received the 2023 Nobel Prize for Medicine
I don’t have a problem with someone evaluating my science abilities if I have the ability to examine theirs.
Let’s do televised qualifying exams for all the political appointees who will be evaluating science grants and proposals. Let’s start with directors of institutes and cabinet secretaries.
How challenging is the prediction of transcriptional responses to CRISPR gene perturbations ?
A simple model, just scaling RNA correlation vectors, results in accurate predictions for many perturbations.
The model is intuitive & grounded in a simple approximation.
1/2
Paper in Nature 150 years ago. Beats reading any single-cell experiment…
