Mikhail Shugay @antigenomics

Can we predict which T cells recognize which peptides? A lot of the signal is hiding in AlphaFold3's hallucinations. enFoldX: instead of one predicted structure, we use the whole noisy ensemble to predict TCR:pMHC binding. 💻github.com/jonlevi/enFoldX biorxiv.org/content/10.648…

KIR Expression Defines a Transcriptionally Distinct CD8+ T-cell Population that Confounds Antigen-Specific T-cell Detection @biorxivpreprint biorxiv.org/content/10.648… 🇩🇰

Glad to share our latest publication in Science! This work reports: 🧬 the first single-cell 3D genome atlas of human tonsil and B cell immunity, 🧬 the first image-based 3D genomics dataset of any human tissue, 🧬 a new function for cohesin loop extrusion, and 🧬 a novel chromatin "curl" structure. Antibody-mediated immunity relies on the generation of point mutations in rearranged immunoglobulin (Ig) loci of activated germinal center (GC) B cells. This process - called somatic hypermutation (SHM) - allows for antibody affinity maturation but also acts at certain non-Ig sites in the genome, thereby contributing to mutations and chromosomal translocations that drive B cell oncogenesis. My collaborator Prof. David Schatz's previous work indicates that SHM susceptibility is controlled by the cooperative action of cis-acting SHM target elements and the architectural properties of topologically associating domains (TADs), but how the genome is spatially organized across multiple length scales as GC B cells develop and activate SHM and how 3D genome architecture influences the targeting of SHM remains unknown. To test the functional requirement of 3D genome organization for SHM, in this work we developed a 3D genome and spatial transcriptome toolbox optimized for clinical tonsil tissue, and used it to define single cell 3D genome architectures and nuclear organization in GC B cells undergoing SHM in normal human tonsil samples and in malignant GC-derived human B cell lymphoma cell cultures. Our new work generated the following key insights: 🧬 At the large scale, the nuclear positioning of TADs is linked to SHM susceptibility, with the nuclear periphery being more permissive to SHM and the nuclear interior being more protected from SHM. 🧬 At the fine scale, increased intra-TAD looping contacts in gene regions are associated with SHM susceptibility. 🧬 Most importantly, through rapid, targeted degradation of cohesin component RAD21, our new results provide the first direct evidence that the cohesin mediated loop extrusion is essential for SHM. We further showed that the effects of loop extrusion on SHM cannot be solely attributed to transcription activity changes. This represents a brand new function of the famous loop extrusion process. 🧬 In addition, we serendipitously discovered a novel chromatin “curl” structure – a chromatin loop with two long (~25 kb) stem regions aligned in parallel with each other (distinct from e.g., CTCF anchored chromatin loops where two stems are aligned in an anti-parallel fashion). To our best knowledge, this is the first report of such a structure outside of the contexts of DNA recombination/transposition. I'd like to thank all my co-authors, especially co-corresponding author legendary immunobiologist Prof. David Schatz, and co-first authors Yubao, Jianshu, and Yuan. It has truly been a wonderful experience working with you. Link to paper: science.org/doi/10.1126/sc…

When comparing clusterings, what makes a better solution? In a new preprint, I describe homogeneity–parsimony curves as a means to quantify the underlying trade-off. Think of it as an ROC curve, but for clustering validation. arxiv.org/abs/2607.20799

@ScienceMagazine definitely muscle)

📢Out now! @houchao1, @YufengShen07 and colleagues show that PLMs predict fitness best when outputs align with evolutionary patterns in homologs, with the peak performance occurring at moderate predicted sequence likelihoods. nature.com/articles/s4358… 🔓rdcu.be/ftBgt

New preprint from the lab! We benchmarked AlphaFold3, AlphaFold2 & other deep learning tools for modeling of 3 immune interface classes: antibody-protein, antibody-peptide & TCR–pMHC complexes. biorxiv.org/content/10.648… #AlphaFold #antibody #TCR @UMDCBMG @BISIumd @UMD_IBBR (1/3)

Excited to share our work on SPEARMINT: peptide–MHC-I binding stability prediction using pLMs! TLDR: We train an assay-aware binding stability predictor that generalizes better to unseen data and show stability complements affinity for immunogenicity and personalized NeoAgs🧵

Great job! Any plans to re-compute existing datasets like 4-donor 10X dCODE and provide some "gold standard" filtered tables (e.g. on huggingface/github)? There is actually a problem with those as multiple groups have processed them with different parameters, and its hard to tell which one is "definitive".

🚨 New preprint! How can we reliably identify antigen-specific T cells from noisy single-cell pMHC multimer experiments? We developed DextraDemixer: a Bayesian framework to disentangle true antigen-specific binding from background noise. doi.org/10.64898/2026.… 🧵 1/7

Excited to share our new paper in @JCI_insight How does genetic risk shape CD4 T cell recognition in #IBD? We linked HLA-DRB1 genotype to >3.1 million memory CD4+ TCRβ sequences in Crohn’s disease, ulcerative colitis and controls.

Excited to share our new preprint @supercs08 : We present TCRDiff, a conditional diffusion model for generating antigen-specific TCRs, moving beyond prediction and toward controllable design of TCRs with specificity and functional potential. Read more: biorxiv.org/content/10.648…

This is an insane paper and I love it arxiv.org/abs/2605.31514

Capturing and Tracking Clonal T-cell Response to Cancer Neoantigens brnw.ch/21x36aN @ChudakovM @antigenomics

My first co-corresponding author paper! Excited to contribute to this work on tracking ultra-rare lymphoma cells using TRUST4, which assembled BCRs for the SMART-seq2 data with the read length of only 35 bp.

Andy Lane @lane_andy

2 months ago

Our lab paper by Ran Xu @Ran47800515 now in print @BloodAdvances . New technology for ultra-rare MRD cell capture for mechanistic discovery: Live-cell Pick-Seq (LiP-Seq): Interrogating ultra-rare mantle cell lymphoma persistent cells after CART19 therapy ashpublications.org/bloodadvances/…

2 4 38 6K 7

Big progress vs cancer, folks. The kind of event curves from randomized trials that we've not seen before for a couple of the most deadly cancers. Congrats to the oncology research community for getting these trial done. #ASCO26, @ASCO

Today we're announcing ESMFold2, an open scientific engine to power prediction, design, and discovery across protein biology. The new model delivers state of the art performance on protein interactions, especially antibodies, a critical modality for therapeutics. We have designed and validated miniprotein binders and single chain antibodies across five therapeutic targets that are important in cancer and immunology. We are seeing very high success rates, and affinities at levels consistent with therapeutic activity. We’re also releasing an atlas of 6.8 billion proteins, and 1.1 billion predicted structures. ESMFold2 is built on a state of the art language model that has been trained on billions of protein sequences. A world model of protein biology emerges through language modeling. We’ve used the techniques of mechanistic interpretability developed to understand large language models to understand the concepts ESM uses to represent proteins. The model’s representation space has a compositional organization of features across scales, levels of complexity, and abstraction, that reflects and mirrors the understanding of protein biology developed through a century of empirical science. This understanding emerges without prior knowledge, just from language modeling of protein sequences. Language models are becoming a powerful substrate to understand and program biology. The design of protein interactions is one of the most fundamental problems in biophysics, and has critical implications for the discovery of new medicines. A simple gradient based search with the model was able to discover high-affinity protein binders. I'm excited by the potential this has to accelerate basic science and the understanding of proteins. And especially for the new avenues it opens up for therapeutic design and medicine.

In a Perspective article, scientists summarize the results of recent studies on how immune systems learn how to target previously unseen variants of pathogens and discuss how statistical physics could help address key unanswered research questions. 📃 go.aps.org/4uPmZzO

Our comment on the urgent need for decentralized scientific databases, particularly in today’s unstable geopolitics, is now online in @NatureGenet. Overdependence on a single authority can jeopardize global scientific access, resilience & continuity. Link: nature.com/articles/s4158…

Our consortium paper on deep profiling of individuals of different ethnicities located across continents just came out. I HPP initiative involving many different labs from aorund the world and an amazing open access resource. Full of cool results. cell.com/cell/fulltext/…