Yuta Nagano
@YutaNotUtah
Followers
62
Following
58
Media
3
Statuses
25
Quantitative immunology, machine learning, medicine MBBS PhD @UCL @UCLMS
Joined June 2021
Excited to announce that our work on #TCR representation modelling is now out on @CellSystemsCP! You can try our model SCEPTR at 🚀.
Nagano et al. present SCEPTR , a TCR language model with a pre-training strategy combining autocontrastive learning and masked-language modeling, to predict TCR specificity, outperforming existing language models and sequence-based methods @andimscience
1
0
4
RT @andimscience: Our work on contrastive learning of T cell receptor representations is now out in Cell Systems! . Give SCEPTR embeddings….
0
20
0
Quant-immuno people: you should know that @andimscience is hiring! As his former PhD student I can attest his group is a perfect mix of brilliant minds with an open, collaborative culture and you'll have an amazing time.
0
1
2
RT @andimscience: How to learn generalizable rules across complex sequence-function maps? 💡In our new preprint we propose a framework for l….
0
13
0
RT @andimscience: Using information theory to study the immune receptor code? 👉 check out this short popular summary of our recent @PNASNew….
0
2
0
RT @andimscience: What are the rules of the immune receptor code? In my talk @KITP_UCSB last Friday I summarised our recent findings, pro….
0
11
0
RT @andimscience: Information theory of the T cell receptor sequence-function map -- our paper now out @PNASNews ! 🔥. How informative (in b….
0
33
0
RT @MVCowley: 🚨 Fresh #preprint 🚨. Targeting #DUX4 protein diffusion when designing #FSHD #therapeutics looks to be a great strategy!. Read….
0
1
0
RT @KaneFos: Very happy to share our latest pre-print studying T cell dynamics in multiple myeloma disease evolution 🧵1/10. .
0
20
0
A huge and sincere thank you to my supervisors @BennyChain @andimscience and co-authors Andrew Pyo @Martina_Milig @jhenderson_sci and Prof. John Shawe-Taylor. It's been great fun and I've learned so much working on this with all of you 🙏💪.
0
0
3
We also find that performance on TCR specificity prediction is not currently bottlenecked by model size.
1
0
2
Our new model SCEPTR is a TCR language model jointly pre-trained on autocontrastive learning and masked-language modelling. We believe autocontrastive learning might help SCEPTR de-convolve VDJ recombination biases from co-specificity signals.
1
0
1
How can we better align protein language model (PLM) pre-training with downstream tasks of interest? For TCR specificity prediction, it turns out that autocontrastive learning is really beneficial! Check out our preprint:
1
2
18
RT @BennyChain: The latest exciting step forward in T cell receptor representation ! @YutaNotUtah outperforms TCR-BERT ! .
0
7
0
RT @andimscience: How to pre-train protein language model to optimize transfer learning? @YutaNotUtah’s PhD work shows prior PLMs struggle….
0
12
0
Really proud and grateful to be part of this amazing work led by @Martina_Milig!.
New preprint!. We investigated the molecular interactions between the TCR CDR loops and find that these may be important in shaping TCR specificity and sequence.
0
0
5
RT @andimscience: Happy to release our preprint introducing an information theory of T cell specificity! . We rank TCR features by their re….
0
19
0
RT @trevoragraham: New preprint: Annie Baker & @BennyChain have led the development of "FUME-TCRseq" a new T-cell receptor seq assay sensit….
0
14
0
Ever dreamed of clustering 10^6 TCR sequences in one second? Or 10^7 TCRs in 10 seconds? Well now you can, thanks to amazing work by @heartnetkung and @andimscience! Check it out:
1
6
21