Lars Velten

Beeswarm plot of the prediction error across different methods of double perturbations showing that all methods (scGPT, scFoundation, UCE, scBERT, Geneformer, GEARS, and CPA) perform worse than the additive baseline.

Line plot of the true positive rate against the false discovery proportion showing that none of the methods is better at finding non additive interactions than simply predicting no change.

Our paper benchmarking foundation models for perturbation effect prediction is finally published 🎉🥳🎉

www.nature.com/articles/s41...

We show that none of the available* models outperform simple linear baselines. Since the original preprint, we added more methods, metrics, and prettier figures!

🧵

Iterative deep learning design of human enhancers exploits condensed sequence grammar to achieve cell-type specificity Yin et al. demonstrate the use of iteratively retrained deep learning models to design synthetic enhancers for progressively higher cell-type specificity. They show the feasibility of model training f...

Interesting piece on automated enhancer design from the Seelig lab: www.cell.com/cell-systems...

Congrats @hoheyn.bsky.social , @lgmartelotto.bsky.social and team!

Scientists make breakthroughs on blood ageing and limb regrowth Two new studies could help develop better therapies for rejuvenation and regeneration

🇬🇧: Finally, English-language coverage by Financial Times, a small London-based newspaper 😉. Thanks again @omarjam.bsky.social and Nahia (@irbbarcelona.org) for making all this possible! www.ft.com/content/a1f8...

Científicos españoles descubren cómo envejece la sangre: un gran hallazgo para la lucha contra la leucemia y el envejecimiento Un estudio desarrollado en Barcelona ha dado con la clave para entender cómo envejece la sangre. El hallazgo podría facilitar la detección temprana de enfermedades como la leucemia y abrir nuevas vías...

😵‍💫: Private TV antenna3 hectically shot an interview 1.5h before their 3pm news, got a lot wrong (no we did not rejuvenate mice by ablating old stem cells, Irv Weissman did that) BUT they made us the good news of the day and it's somehow funny www.antena3.com/noticias/cie...

😬: SER Catalunya, most-listened radio around here, interviewed me at 8:40 (instead of 8:50)... they reached me in the elevator, put me live on air and started asking in Catalan. Had to guess what they meant and improvise. The good thing, I was awake after this 😅 (can't upload audio to bsky - 7/n)

Dès l’âge de 50 ans, les cellules de notre sang s’appauvrissent Une équipe espagnole a retracé l’évolution des lignées de cellules du sang chez 13 personnes de 35 à 70 ans. A partir de la cinquantaine, le nombre de cellules souches capables d’engendrer des cellule...

🇫🇷: @scherermich.bsky.social was brave enough to speak to @lemonde.fr in French 💪! Excellent article that definitely sticks out with the accompanying visual.
www.lemonde.fr/sciences/art... (6/n)

„Blutbarcodes“ könnten Alterung verlangsamen Auch das Blut altert: Ab 50 Jahren verringert sich die Vielfalt der Stammzellen, die Blut produzieren. Ein Forschungsteam hat nun ein „Barcodesystem“ entwickelt, das die verschiedenen Blutstammzellen ...

🇦🇹: Austrian ORF covered both the blood ageing context and the DNA methylation discovery in two minutes on the radio after interviewing @martinabraun.bsky.social and me. Real quality science journalism. science.orf.at/stories/3230... (5/n)

"A partir de los 50 casi todas las células madre se llaman Rodríguez" (Y eso es malo) Un estudio liderado por dos centros de investigación de Barcelona ha descubierto que, a partir de esa edad, cada vez tenemos menos células madre sanguíneas y las que hay se replican cada vez con más e...

😅: "From age 50 almost all stem cells are called Rodriguez (and that's bad)". All credit for this to @alejofraticelli.bsky.social - family names are a fantastic analogy to explain clonal processes cadenaser.com/nacional/202... (4/n)

Telenotícies - Envelliment de la sang: d'un ecosistema ric i variat a un de dominat per unes poques cèl·lules - 3Cat Una investigació catalana rastreja els "codis de barres" escrits a l'ADN i revela què passa a la sang quan envellim.

❤️: Catalan public TV 3cat nicely covered the science and the emotions when making a cool discovery in their evening news (almost like Sports reporting) www.3cat.cat/3cat/envelli... (3/n)

First of all, here's the science, in case you missed it. bsky.app/profile/lars...
Now come the media highlights (2/n)

"A partir de los 50 casi todas las células madre se llaman Rodríguez" (Y eso es malo) Un estudio liderado por dos centros de investigación de Barcelona ha descubierto que, a partir de esa edad, cada vez tenemos menos células madre sanguíneas y las que hay se replican cada vez con más e...

Thanks to phenomenal work by @crg.eu and @irbbarcelona.org press offices, we got to communicate our findings on 🧬 methylation barcodes and 🩸 blood ageing to a wide audience. A cool experience!

🧵 with highlights in 5 languages and funny moments 😬😅

cadenaser.com/nacional/202... (1/n)

Single-cell ultra-high-throughput multiplexed chromatin and RNA profiling reveals gene regulatory dynamics - Nature Methods This work presents SUM-seq, an ultra-high-throughput method for co-profiling chromatin accessibility and gene expression in single nuclei across multiplexed samples, advancing the study of gene regula...

SUM-seq out @natmethods.nature.com !


🚀 Ultra-high-throughput Multiplexed snATAC+RNA

Used to:
⏳ link temporal macrophage GRNs to immune disease genetics
🩸 map T cell regulatory landscapes
🧬✂️ dissect TF function in hiPSC differentiation via CRISPRi/a screens
doi.org/10.1038/s41592-025-02700-8
🧵

Thank you Jorge!

What a year 😅 really really happy that both these projects worked out better that we could ever have planned, and the Barcelona scientific environment was super important for both of them!

It was great to see how this story rapidly developed from a serendipitous discovery on DNA methylation to a project with relevance to blood stem cell ageing and beyond. Big thanks to the funders, especially EHA and AECC.

With help of @leifludwig.bsky.social @caleblareau.bsky.social & colleagues, we compared this to mitochondrial lineage tracing; TL;DR: mitochondrial genetics is complicated (12/n)‬

Functional phenotyping of genomic variants using multiomic scDNA-scRNA-seq Genomic variation ranging from single nucleotide polymorphisms to structural variants can impact gene function and expression, contributing to disease mechanisms such as cancer progression. The system...

With help of Lars Steinmetz and Dominik Lindenhofer, Chelsea Szu-Tu added RNA-seq capabilities to the protocol, based on their SDR-seq www.biorxiv.org/content/10.1... (11/n)

@biotechpedro.bsky.social showed that the same CpGs that work in blood also work on endothelia, suggest a broad applicability to other tissues! (10/n)

@martinabraun.bsky.social showed that in human a) clonal expansions are ubiqutious by age 50 and b) CH mutation linked clones are just a small subset of clonal expansions, with similar functional biases as “driverless” clones (9/n)

Depleting myeloid-biased haematopoietic stem cells rejuvenates aged immunity - Nature Antibody-mediated depletion of myeloid-biased haematopoietic stem cells in aged mice restores characteristic features of a more youthful immune system.

@indrasingh.bsky.social showed that in mouse, some HSC clones with poor ability to transplant grow to a large size, whereas many young-like clones persist until old age – fueling ideas on rejuvenation therapies proposed by Irv Weissman last year: www.nature.com/articles/s41... (8/n)

We then built an algorithm, EPI-Clone, that extracts clonal information from scTAM-seq data. It recapitulates ground truth clonal labels with high accuracy. This was more or less our preprint, online since April 2024. Since then a lot has happened: (7/n)

To make this observation, it was essential to measure methylation at single cell, single CpG resolution. Our earlier method scTAM-seq does exactly that, is straight-forward to run on @missionbio.bsky.social , scales to 10,000s of cells, and is compatible with CITE-seq: doi.org/10.1186/s130... (6/n)

This made us wonder if these two layers can be separated. Again to our surprise, we discovered that the CpGs that change with differentiation (“dynamic” CpGs) are different from the ones that are clone-specific (“static CpGs”) (5/n)

We performed targeted single-cell profiling of 453 CpGs on cells that had been barcoded with a lentivirus, so we knew clonal identity. To our big surprise, these data clustered both by differentiation state, and by clone. (4/n)

Our tissues are composed of clones. Lineage tracing is key to understand tissue biology, but requires genetic engineering. In human, somatic mutations contain lineage info, but are too spares to be read out effectively. What if there was a heritable mark on DNA that mutated much more rapidly? (3/n)

This project was a wild ride with @alejofraticelli.bsky.social + team. At its heart is an unexpected discovery we stumbled on by chance… Great work by @scherermich.bsky.social 🧬🔨, @indrasingh.bsky.social 🐭🩸, @martinabraun.bsky.social 👤🩸and Chelsea Szu-Tu 🧬💪 (2/n)

Clonal tracing with somatic epimutations reveals dynamics of blood ageing - Nature The discovery that DNA methylation of different CpG sites can serve as digital barcodes of clonal identity led to the development of EPI-Clone, an algorithm that enables single-cell lineage tracing th...

Out @nature.com: Clonal tracing with somatic epimutations

🧬 Single cell methylome encodes cell state & clonal identity

🔨 EPI-Clone reads out both (+mutations, +RNA) at scale

🩸 Clonal expansions of HSCs are universal from age 50, not driven by CH mutations

doi.org/10.1038/s415...
🧵

Super happy and proud of this one. We and others have developed tools for dimensionality reduction for many years. MOFA (linear) and other non-linear generalisations, space/time. scDORI fuses this with GRN inference. It’s a nice and useful piece of software. Here applied to GBM biology.

LoxCode in vivo barcoding reveals epiblast clonal fate bias to fetal organs Much remains to be learned about the clonal fate of mammalian epiblast cells. Here, we develop high-diversity Cre recombinase-driven LoxCode barcoding…

A triumph of perseverance from twitterless Tom Weber, Christine Biben and the team, our in vivo barcoding "LoxCode mouse" used to resolve epiblast fate to fetal organs is finally published in @cellpress.bsky.social and available through @jacksonlab.bsky.social www.sciencedirect.com/science/arti...