Constantin Ahlmann-Eltze

It definitely helps, but I think it's also simply that most scientists have a LinkedIn account and occasionally check it, whereas still only a minority has a bluesky account.

Yeah, the amount of engagement on LinkedIn compared to bluesky is crazy. I recently posted about a new paper and it got 480 likes on LinkedIn vs 7 on bluesky (and I have more followers here than there...) :/

We wrote a review of representation learning methods of single-cell RNA-seq data, where we compare factor models, autoencoders, contrastive learning, and foundation models 🎉

rnajournal.cshlp.org/content/earl...

Postdoctoral Researcher or Senior Scientist (AI × Biology)

Boeva Lab is #hiring postdocs and senior scientists in AI/ML for Cancer Biology! Check the offer and apply at: jobs.ethz.ch/job/view/JOP...

View from the hotel room

Poster session 2024, with Valentina Boeva, Constantin Ahlmann-Eltze and others

Wednesday afternoon hike incl. swim in the mountain river

Another view from the hotel room

Apply for the Ascona workshop "Statistical and AI methods for multi-modal multi-scale modeling of biological systems", 28 Jun-3 Jul 2026 on Monte Verità, Lago Maggiore at the foot of the Swiss Alps.

ascona2026.sciencesconf.org

Save the date: April 9 from 4pm to 6pm CET. Our department is hosting an online seminar with @noeliaferruz.bsky.social @sdomcke.bsky.social @const-ae.bsky.social who will talk about models for protein design, large-scale perturbation screens, and benchmarking of perturbation prediction models.

Join me in 5.5h, when we discuss our benchmark of perturbation prediction models and what the right metric is to assess if a gene expression prediction is good!

🚀 Excited to share our new preprint: msBayesImpute - A Versatile Framework for Addressing Missing Values in Biomedical Mass Spectrometry Proteomics Data
👉 Improves imputation accuracy, normalization, and differential expression detection
📝https://www.biorxiv.org/content/10.1101/2025.10.02.679746v1

Heads up: ignore samtools dot org, similarly minimap2 dot com and likely others. It's owned by a known phishing site and while the binaries they offer look valid currently (but note they may be serving us different binaries to others), that could change.

Ie: it's not us (Samtools team)! Be warned

We're excited to share that our preprint on anndataR, a new package bringing Python's AnnData to R, is now available on bioRxiv 🎉

🔗 Read the paper: www.biorxiv.org/content/10.1...
💻 Check the package in action: anndatar.data-intuitive.com

An arrow with a LaTeX equation

Trigonometric functions and a unit circle

A bivariate change model with structured residuals

A hierarchical model of cognitive abilities

Now on CRAN, ggdiagram is a #ggplot2 extension that draws diagrams programmatically in #Rstats. Allows for precise control in how objects, labels, and equations are placed in relation to each other.
wjschne.github.io/ggdiagram/ar...

Makes sense. I imagine this would simply be my primary use case, and I would prefer not having to refer to `x` twice. Something like:

replace_values <- function(x, ..., from=NULL, to=NULL, lookup=NULL){
if(! is.null(lookup)){
from <- names(lookup)
to <- lookup
}
...
}

This looks great! I think it would be neat if the functions also supported named look-up vectors in addition to the `to` and `from` arguments :)

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!

🧵

Pre-Cancer Immunology The Pre-Cancer Immunology Lab (James Reading Lab) is mapping pre-invasive T cell dynamics during carcinogenesis to detect and intercept cancer development.

🚨 PhD Position available in our lab 🚨 exploring the power of blood immune multi-omics to detect lung cancer years prior to clinical diagnosis in a unique cohort of >10,000 CT screened individuals.
✅ Wet & dry lab
✅ September 2025 enrolment
✅ UK tuition fees only

www.ucl.ac.uk/medical-scie...

I wrote about AI foundation models for biology last year: www.nytimes.com/2024/03/10/s... Benchmarking tests since then aren't finding that they're better than simpler models of how genes and cells work.

Deep-learning-based gene perturbation effect prediction does not yet outperform simple linear baselines - Nature Methods The analysis presented in this Brief Communication shows that, despite their complexity, current deep learning models do not outperform linear baselines in predicting gene perturbation effects, thus e...

An analysis shows that current deep learning models do not beat linear baselines in predicting gene perturbation effects, thus emphasizing the importance of further method development and evaluation. @const-ae.bsky.social @wkhuber.bsky.social @s-anders.bsky.social

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

Haha, would also be a succinct summary of most of my academic work 😅

Deep-learning-based gene perturbation effect prediction does not yet outperform simple linear baselines Nature Methods - The analysis presented in this Brief Communication shows that, despite their complexity, current deep learning models do not outperform linear baselines in predicting gene...

And lastly, a big shout-out to @wkhuber.bsky.social and @s-anders.bsky.social!

Link to pdf: rdcu.be/ey7x0

- Li et al. doi.org/10.1101/2024.12.23.630036
- Chen Li et al. doi.org/10.1101/2024.12.20.629581
- Wong et al. doi.org/10.1093/bioinformatics/btaf317

And probably many more that I am missing here.

- @kasparmartens.bsky.social et al. openreview.net/forum?id=eb3ndUlkt4
- Gaudelet et al. doi.org/10.48550/arXiv.2404.16907
- @aaronwtr.bsky.social et al. openreview.net/forum?id=t04D9bkKUq
- Bendidi et al. doi.org/10.48550/arXiv.2410.13956
- Wu et al. doi.org/10.48550/arXiv.2408.10609

I also encourage everyone to checkout the parallel efforts by groups around the world who came to similar conclusions:
- @ekernf01.bsky.social et al. doi.org/10.1101/2023.07.28.551039
- Csendes et al. doi.org/10.1186/s12864-025-11600-2
- @kasia.codes et al. doi.org/10.1186/s13059-025-03574-x

*We benchmarked scGPT, scFoundation, GEARS, CPA (which claim predictive ability), and scBERT, Geneformer, and UCE (which do not claim this ability). I can't comment on methods released in the last 5 months. For a summary of recent developments, see @ekernf01.bsky.social's post

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!

🧵

In October 2024, I twote that "something is deeply wrong" with what we now call virtual cell models. A lot has happened since then. How am I updating? New blog post: ekernf01.github.io/virtual-cell...

That's amazing. Congratulations 🎉

Tracing colorectal malignancy transformation from cell to tissue scale The transformation of normal intestinal epithelium into colorectal cancer (CRC) involves coordinated changes across molecular, cellular, and architectural scales; yet, how these layers integrate remai...

Going from methods dev to full on cancer bio has been tough. All the more excited to see this out at last—

www.biorxiv.org/content/10.1...

@pascual-reguant.bsky.social
@brukerspatial.bsky.social
@hoheyn.bsky.social

With {tidylog}, you see friendly messages like: summarise: now 3 rows and 3 columns, ungrouped filter: removed 2 rows (67%), one row remaining

With {tidylog}, you see friendly messages like:

summarise: now 3 rows and 3 columns, ungrouped

filter: removed 2 rows (67%), one row remaining

✨ It’s like having a gentle co-pilot, helping you track what each step is doing — super useful for debugging and learning!

Spatialproteomics orchestrates workflows to analyze highly multiplexed images. It segments cells, processes images, quantifies proteins, predicts cell types, and provides neighborhood analysis methods, all while integrating into the scverse ecosystem.

New preprint out!

We introduce 𝐬𝐩𝐚𝐭𝐢𝐚𝐥𝐩𝐫𝐨𝐭𝐞𝐨𝐦𝐢𝐜𝐬, a Python package for end-to-end processing and analysis of highly multiplexed immunofluorescence imaging data.

Built on xarray and dask, with seamless integration into the scverse ecosystem.
www.biorxiv.org/content/10.1...

(2/3)

Constantin Ahlmann-Eltze showcased impressive work on assembling an atlas of T cells in precancerous samples. They developed the R packages treelabel and Shinytreelabel, which showed GITR+ Tregs are enriched in several precancerous samples, suggesting avenues for future treatments