@UDemael profile picture
Ursule Demaël @UDemael

🦠 PhD candidate @UCL & @MGHNeurosurg 🦠 // co-md & venture fellow @NucleateUK// @join_polaris //previously @UniofOxford & @HOLOBIONT_ //Immunology & Biotech ✨

Joined March 2021
  • Tweets 130
  • Following 733
  • Followers 250
23 Photos and videos
Ursule Demaël retweeted
Kyle Duffy@kyduffy
31 Jul 2025
Cohere just launched its first multimodal LLM. My team and I built it—check it out! cohere.com/blog/command-a-vi… It beats OpenAI GPT 4.1, Meta Maverick 4, Pixtral Large, and Mistral Medium 3 models on most important benchmarks.

Introducing Command A Vision: Multimodal AI

Command A Vision excels across enterprise image understanding tasks while keeping a low compute footprint.

cohere.com
2
6
49
2,671
Ursule Demaël retweeted
Matthew V. Cowley@mvcowley
21 Mar 2025
Less than a week left to submit abstracts for the 3rd London Quantitative Immunology Day! qimmuno.com/ldnday/. Register and share your work with a diverse set of biologists, physicists, mathematicians all united by a common interest in immunology. @UDemael
4
3
703
Ursule Demaël retweeted
Matthew V. Cowley@mvcowley
22 Dec 2024
As a happy end of year post 🌲🎆 I'm delighted to share I've been awarded a @RoyalFreeChty ECR grant! I'll be studying the clonality of T cells in fascioscapulohumeral dystrophy. Huge thanks @BennyChain, @iit_ucl, and collaborators at Queen's Square and King's College. 🙏
1
1
14
707
To anyone looking for a PhD in computational immunology, check this project out! Andreas leads an incredible group and is an amazing supervisor!! ⭐️ @andimscience qimmuno.com/openings/

1
1
4
3,534
239
Ursule Demaël retweeted
Nucleate@NucleateHQ
21 Oct 2024
Join us on the latest Nucleate Podcast with Luke Timmerman (@ldtimmerman), award-winning biotech journalist and founder of @timmermanreport! Listen to the full episode on Signal: signal.nucleate.xyz/biotech-…
2:19
1
2
5
395
Ursule Demaël retweeted
Andreas Tiffeau-Mayer@andimscience
9 Oct 2024
Information theory of the T cell receptor sequence-function map -- our paper now out @PNASNews ! 🔥 How informative (in bits) is the α or β chain? When does partial information limit predictions? insights into synergy, Renyi entropy, optimal compression & more 👉
https://www.pnas.org/doi/10.1073/pnas.2408696121

ALT https://www.pnas.org/doi/10.1073/pnas.2408696121

2
31
132
13,692
Super excited to release this short movie about our group, the Stauss Lab at UCL's Institute of Immunity and Transplantation🧪🔬 Our research aims to engineer T cells against cancers and autoimmune diseases Huge thanks to Charlie Yuille and his team at CHIK for the film!
IIT@iit_ucl
30 Sep 2024
Three years today since we moved into the Pears Building! 🎊 To mark this, we've just launched a new film celebrating some of the fantastic scientists from the Stauss Lab here at the Institute 🎬 buff.ly/3XKAEZV #Immunotherapy #TCells #CancerResearch #AutoImmuneDisease
2:19
15
498
Ursule Demaël retweeted
Allegra Petti@allegra_petti
23 Sep 2024
Join us at MGH/HMS for a postdoc! We’re recruiting computational biologists to analyze scRNA-seq, spatial transcriptomic, and bulk genomic data from diverse cancer types. More info here: pettilab.mgh.harvard.edu/hom… Please RT!
1
1
402
Had an incredible time at the Nucleate 2024 Summit in Santa Cruz (CA)🌞🌲 @NucleateHQ Learnt so much from incredible keynotes including Rick Klausner from @NCI @Altos_Labs (and many more), @tkalil2050 , Alexander Titus at @DeptOfDefense and @ilangur from @ARIA_research
1
12
668
more replies
This closes months of a VC fellowship where 5 of us have interviewed and diligenced the top of the 162 biotech companies participating in the @NucleateHQ Activator. @kulesatony and @jbpin90 were incredible mentors in this process!
1
2
106
If you're interested in getting in touch, supporting, or investing in our brilliant Activator companies, reach out! ...as always Nucleate events always leave me so proud of the people I get to work with- massive congrats to the whole team for an excellent Summit @NucleateHQ
2
91
Ursule Demaël retweeted
Andreas Tiffeau-Mayer@andimscience
11 Jun 2024
How to pre-train protein language model to optimize transfer learning? @YutaNotUtah’s PhD work shows prior PLMs struggle to predict TCR specificty & uses contrastive learning to overcome this limitation. 👉arxiv.org/abs/2406.06397

Contrastive learning of T cell receptor representations

Computational prediction of the interaction of T cell receptors (TCRs) and their ligands is a grand challenge in immunology. Despite advances in high-throughput assays, specificity-labelled TCR...

arxiv.org
4
12
50
6,865
Ursule Demaël retweeted
Benny Chain@BennyChain
29 May 2024
Martina's elegant new paper defines an important and under-appreciated role for molecular interactions between CDR loops of the TCR !
bioRxiv Immunology@biorxiv_immuno
29 May 2024
Intra- and inter-chain contacts determine TCR specificity: applying protein co-evolution methods to TCRαβ pairing biorxiv.org/cgi/content/shor… #biorxiv_immuno
7
28
3,235
Purity is also not the criteria by which scientists judge "good" science "If we study the criteria that underlie a scientist's own valuation of science, we shall certainly not find purity among them"...
1
99
more replies
...one last snippet I loved is him marvelling at the diversity of function of biological macromolecules: "They insulate, they fill out, they fetch and carry, they prevent the Organism as a Whole from falling apart or from dissolving in water, they prop up, they protect..."
1
1
95
"...they attack and defend, they store energy and catalyze its transfer, they store information and convey messages, and sometimes they themselves are messages!"
1
91
...we have developed a thwarted view of the superiority of "pure" research detached from intended use or application The view is that useless = good
1
76
more replies
"Pure Sciences are best not because they lack practical application" but because they are not acquired by error-prone empirical methods pure ≠ useless "Useful Arts are "less noble" than the Pure Sciences-but because of their fallibility, not because of their usefulness"...
1
79
Load more