Theofanis Karaletsos
@Tkaraletsos
Followers
4K
Following
8K
Media
67
Statuses
2K
Head of AI @cziscience | https://t.co/TIpJn8YQlg | probabilistic and deep ML
Joined December 2013
RT @freethinkmedia: What if AI virtual cells could help us decode disease, test medication, and rewrite biology? . Read the latest edition….
0
27
0
RT @thjashin: Autoregressive models are too restrictive by forcing a fixed generation order, while masked diffusion is wasteful as it fits….
0
55
0
Proud to share a big step toward building systems that help us simulate and predict the behavior of cells!. GREmLN is a graph-based transformer model, designed to reason across biological data by embedding gene regulatory network architecture directly into its attention.
1
2
10
Fun write up by @ElliotHershberg on virtual cells.
What Are Virtual Cells?. Two months ago, I started working on an essay to answer this question. The goal was to cover a few of the recent research results. Simple as that!. Instead, I went down a rabbit hole exploring ideas around cellular simulation
1
0
4
ICML 25 paper on variational guidance for diffusion models accepted. Happy to share that our diffusion model guidance paper with @farrinsofian, @kpandey008, @felixDrRelax, and @StephanMandt on casting control for guidance as variational inference with auxiliary variables was.
🚀 News! Our recent #ICML2025 paper “Variational Control for Guidance in Diffusion Models” introduces a simple yet powerful method for guidance in diffusion models — and it doesn’t need model retraining or extra networks. 📄 Paper: 💻 Code:
1
7
65
Is this the greek version of Wrexham?.Go @kostascrypto !.
⚽️ Just backed a small football team from the birthplace of Aristotle in 🇬🇷 . in a lower division today, but aiming for the national leagues, and hence why they do need modernization!. With support from Sui, we’re bringing blockchain to the pitch: on-chain stats, NFTs,.
1
0
4
Of course a deeper error analysis here will eventually reveal what actually works and why, and whether it captures perturbations or just basal states very well, but I really enjoy seeing these little results and explorations.
0
0
2
Adding our model overview again:. 2/4.
chanzuckerberg.com
Use it to predict cell types, disease states, gene-gene interactions via prompting, and more.
1
0
1
Thanks to the @tahoe_ai hackathon and the amazing team that did the OODyssey work I learned sth new about our recently released Transcriptformer model: .It did well in the wild on disease-relevant, unseen (in this case cellular drug perturbation) data that they use for their drug.
3
1
19
RT @cziscience: TranscriptFormer is a new AI model trained on millions of cells from 12 different species, spanning 1.5 billion years of ev….
virtualcellmodels.cziscience.com
TranscriptFormer is a family of generative models representing a cross-species generative cell atlas trained on up to 112 million cells spanning 1.53 billion years of evolution across 12 species....
0
43
0
Proud to share TranscriptFormer - a generative model for single-cell transcriptomics, trained on 112 million cells across 12 species spanning 1.5 billion years of evolution. TranscriptFormer is distinct in being both generative and multi-species. Built entirely on raw gene.
2
12
105
RT @nalidoust: Meet the judges & mentors for the Tahoe Deep Dive Hackathon & thanks @awscloud for offering the prize: $25K AWS credits to w….
0
12
0
RT @zavaindar: We're launching ACHIRA, a newco DIMENSION cofounded alongside @jchodera & @tkaraletsos. We're building foundation simulation….
0
21
0
I also want to point out how exciting it is to be part of two concurrent revolutions in AI in the life sciences between the virtual cell work we pioneer at @cziscience and the AI for atomistic simulations for drug discovery with Achira. Chemistry and biology are inextricably.
4
0
10