The IMMREP25 TCR specificity prediction challenge is now live on Kaggle! Get involved: https://t.co/7Jqbt8KhbX Huge thanks to: Matthew Noakes, Damon May, @EveRichardson20 @bpeters @AdaptiveBiotech @iedb_ @airr_community

Participated internally last year and loved it (though was humbled by how difficult it is to build a generalisable model).

If you’re interested in prediction using lots of lovely high-dimensional multiomic data - I highly recommend participating in this public challenge:

@prarija @SagidaBibi, Flo McLean, Lisa Schimanski, @prarija, Marie Ghraichy, Valentin von Niederhäusern, @jtrueck, @LizClutterbuck, @DanScholar, Kerstin Luhn, Alain Townsend, @bpeters, @ajpollard1 and senior authors Charlotte Deane&Dominic Kelly!

@prarija Thrilled this work is out- many thanks to my many fab co-authors on this work, to the reviewers and editor, and to the participants.

@prarija More generally- our database of ~1000 Ebola mabs from literature, in conjunction with our simple seq identity search tool CloneSearch, could identify at least one of top ten clonotypes in 2/3 of our cohort as EBOV-specific.

From an EBOV vaccination standpoint, most interesting that IGHV3-15 cropping up time and time again, incl. here in monoclonals isolated from vaccinee plasmablasts which target the receptor-binding region of EBOV GP (@prarija).

BCR repertoire sequencing in 40 Ad26.ZEBOV/MVA-BN-Filo vaccinees in work with @OxVaccineGroup:

Thanks to the organisers for a fantastic conference so far and an opportunity to talk about our work at @iedb_ @ljiresearch

Thank you to my co-authors Raphael Trevizani, J Greenbaum, Hannah Carter, Morten Nielsen and @bpeters, the @iedb_, as well as the Editor and reviewers.

In short - PR-AUC represents classifier performance on a specific dataset, ROC-AUC reflects classifier performance. Great related work by Matthew McDermott & colleagues https://t.co/jQLWT6BUiO.

Unless your imbalance is tied to the score distribution - i.e adding more negative data would make the problem easier - in which case the ROC curve reflects an actual difference in problem difficulty (e.g. including FWR residues in paratope prediction).

This false statement has become widespread - and it’s a real shame, given the beauty and convenience of the ROC curve. We show plainly via simulation that this is False.

Read our (@bpeters) article in Patterns explaining why contrary to recent popular belief, the ROC-AUC does not inflate performance estimates in imbalanced data 👇

New from the @iedb_ team at LJI: "Next-generation IEDB tools: a platform for epitope prediction and analysis" https://t.co/KIT6QnfXkR #DataScience #Immunology #ImmuneSystem #BigData #Epitopes #CompBio @SetteLab @bpeters

I am excited to announce our paper predicting #immune responses after B. pertussis (#whoopingcough) booster vaccination based on the pre-vaccination immune state is coming out in the March 25th issue on @CellRepMethods https://t.co/tpc1qgxGFA #SystemsBiology #Bioinformatics (1/8)

#MucosalImmunologists and #EvolutionaryImmunologists: where does J chain come from? Check the latest piece by the Flajnik lab

For those interested in advancing B cell phylogenetics with single cell data, my lab @dartmouth is recruiting postdocs, programmers, and grad students. Postdoc already posted, but feel free to contact me about others: https://t.co/jvD57153g0.

OPIG researchers @mijr12 and @AlissaHummer are hosting a conference on Adaptive Immune Receptors: structural modelling & immunoinformatics. 5th April 2024 at University of Oxford Conference website, registration, abstract submission: https://t.co/2Zi04IpdTy Confirmed speakers:

On AIRR 15: Germline databases or adventures into the allelic underworld with Dr. Corey Watson and Dr. William Lees