Profile avatar
alexr.bsky.social
personalized cancer immunotherapy = genomics + immunology + machine learning + oncology (pirl.unc.edu)
1,022 posts 5,534 followers 949 following
Regular Contributor
Active Commenter
Posts 15 Comments 35

Boltz-1 and the Future of Biomolecular Foundation Models, seminar by @gcorso.bsky.social & @jeremywohlwend.bsky.social 🧬🧶 youtu.be/K-gzTJMy1ag?...

submitted 91 days ago • 0 comments

Any tips for running Boltz across 1000+ 3-6 chain inputs from the depths of GPU poverty (aka a single machine with an A100)?

submitted 1 hour ago • 2 comments

Gradient of antigen personalization: 1) fixed set of targets 2) select patients-relevant targets from fixed set based on biomarker 3) discover & synthesize novel targets per patient

submitted 1 day ago • 0 comments

Convinced @sergeyf.bsky.social to run Boltz for me, slowly you will all become biologists

submitted 1 day ago • 0 comments

bsky has @dereklowe.bsky.social, we can win this thing

submitted 1 day ago • 1 comment

If you generate an AI image and then feed as the input to generate another one, actually the process always converges in this fixed point. No one knows why.

submitted 1 day ago • 0 comments

Where can I find a digitized form of the Byblos syllabary?

submitted 1 day ago • 0 comments

The only uses for AI that have really stuck for me are: (1) generating Python plotting code with Claude & o3 (2) literature search and research brainstorming with ChatGPT Deep Research / o3 & Future House tools Both extremely high value! (Broader vibe coding usually a time sink)

submitted 2 days ago • 2 comments

I really don't agree. I really think that hoping that a space not filled with Nazis would have a nicer community vibe rather than being abusive and abrasive is not, I dunno, some kinda failing, it's a good thing to want and to strive towards. We spend our time here, why not want it to be pleasant?

submitted 3 days ago • 6 comments

As mass firings decimate US federal agencies, what does it mean for the FDA to phase out animal testing requirements? My latest for The Scientist.

submitted 5 days ago • 2 comments

It’s time: @benjamingvincent.bsky.social and I are going to do a cancer immunotherapy / cool research / accelerating future medicine podcast. Any tips for getting started?

submitted 11 days ago • 1 comment

I think this poster at AACR may have flown under the radar, but Tempus has a great dataset for (mostly) unbiased estimation of prevalence of rare fusion driven cancers and for NUT carcinoma they found that ~2/3 of cases are misdiagnosed: aacrjournals.org/cancerres/ar...

submitted 12 days ago • 0 comments

Friday futureposting: “concierge” medicine is surprisingly low in ambition, it’s still rare to develop a new therapy even with nearly unlimited resources. As we sink deeper into techno-oligarchy, motivated patronage will let us speedrun glacial therapeutic development timelines.

submitted 12 days ago • 1 comment

We're still waiting for another round of experiments to fully resolve this but here are some preliminary results: {mRNA, peptide + CpG} >>> {peptide + polyIC, peptide + montanide} STING agonist does not significantly enhance CpG Upcoming data with poly-ICLC +/- montanide

submitted 13 days ago • 3 comments

Riff Cohen great for playing “guess the language” with my kids (Is she playing any shows in France this summer?)

submitted 13 days ago • 0 comments