Pierre-Paul Axisa

Out today in Nature Genetics: “Quantification method affects replicability of eQTL analysis, colocalization and TWAS,” led by outstanding postdoc @taylorhead.bsky.social at @mdanderson.bsky.social. This was a full lab effort, and I'm proud of them all for their hard work! rdcu.be/a2jiWvAMFaVa

We spend so much time talking about “the future” of scientific discovery, but why not the future of the people responsible for making those discoveries What does it mean to “accelerate innovation” if we aren’t equally committed to expanding opportunity?

The kids are alright

What do you think about AI generated text detection tools? I don’t know much about it practically. Reliable enough? Too costly anyway? Text can be partly AI generated but have sound reasoning anyway? Do you think it could be useful eg for random screens or when readers flag content?

#academiclife

and completely opaque transformations to the data, and cannot be relied upon for almost anything. These extremely valuable datasets are often obtained at great costs and deserve that we take their analysis extremely seriously. We are not going to understand what these data tell us 13/n

The purpose of this thread is not to shit on the work of others but to defend our approach of being extremely careful and deliberate about methods. We are really concerned about typical standards in the single-cell omics field and the endless proliferation of methods that apply complex, ad hoc, 12/n

Hey, I heard about Bonsai, but now I heard there is another method called MILK that also visualizes scRNA-seq on trees. And it is much faster! Should I just use that then? A tree is a tree, right? TLDR: No. #methodsmatter Not even close. On realistic synthetic data MILK performs worse than UMAP.

Interdisciplinary research is great and necessary, and also has consistently lower funding success www.nature.com/articles/nat... #academiclabourmarket #academia #research

Damned it got me right in the intrinsically disordered region

I wrote a post about how slice_sample() and other slice_* functions work in #tidyomics #Bioconductor tidyomics.github.io/tidyomicsBlo...

Since scRNAseq happened I have heard many times "we don’t need large sample sizes, we have SiNgLe cells"

I have recently noticed a string of vandalism on our floor. These outrageous acts seem to be correlated with visits from a certain corn geneticist @jrossibarra.bsky.social

Two weeks ago the @economist.com had a leader "In praise of designer-ish babies". I wrote a letter to the Editor about this, explaining why the UK system is so confident that this should remain unlawful. The Economist has kindly published a shortened form+ my full letter below lnkd.in/p/ef-n78N8

I am getting dizzy with all the flashy LLM generated diagrams and charts in scientific communication. I want to understand what your research is about, a clear title and an odd box and arrows will do perfectly well. I remember nostalgically my postdoc days where lab meetings were chalk only.

Want to explore and understand the structure of your single-cell omics or other high-dimensional data? Try Bonsai! www.nature.com/articles/s41... Check out our tutorial videos with 1. Highlights of the Bonsai-scout exploratory analysis tool: www.youtube.com/watch?v=TE7T...

How do you visualize highly complex data without losing important information? Prof. Erik van Nimwegen’s team developed a new tool, Bonsai, that provides a faithful picture of the structure in the data. @erikvannimwegen.bsky.social @unibas.ch @dhdegroot.bsky.social @imbavienna.bsky.social

1/7 Prestige sucks. Not because recognition is bad. But because science has spent decades confusing journal prestige with scientific quality. And that shapes what gets published, what gets rewarded, and ultimately what science gets done.

As promised, we now share some details why it took 32 months to get this Matters Arising published, and, since many asked, share some comments on the reply to our critique. All details can be found in this small FAQ vincentholst.github.io/the_curious_.... Below is a summary 🧵

I'm attending online event about a #NIH competition to develop a data-sharing index (S-Index), but am surprised that there aren't more people here (it's hybrid). www.nei.nih.gov/sindex Nice idea for NIH to have a competition for developing a S-index; we've just heard about the winners. 1/n

9/11 This case shows how, with complex data from large surveys, small errors in distance treatment or distortions can produce apparently significant signals. Science works exactly like this: strong claims are checked and, if necessary, corrected.

Many of us profs are struggling w/ AI use by students. But a source of frustration I didn't anticipate: AI use by colleagues. Because if I had wanted an AI’s opinion, I too can prompt ChatGPT 🙄 More in my reflection on "Artificial Intelligence among University Professors" jean.fan/2026/07/16/a...

Very good piece by @tuuliel.bsky.social @nevillesanjana.bsky.social @singht.bsky.social & colleagues: “Current challenges in GWAS integration and fine-mapping for variant interpretation” 🧪🧬 www.biorxiv.org/content/10.6...

As a Gilliam Fellow, it’s great to see @hhmi-science.bsky.social continue pushing toward evaluating science on its scientific merit rather than the prestige of the journal it’s in. The future of science should be open, accessible, and focused on the quality of the work, not the name on the cover.

"IF is a flawed metric...not an indication of the quality of published research...But, we are not naive. We know IF influences authors' decisions about where to publish". ☝️ The dilemma every journal faces. And that last sentence is a huge understatement. 1/2 journals.biologists.com/dev/article/...

There's so much to say about this, but what is particularly striking is that the keynote speaker said absolutely nothing about using Claude to write his academic work until I intervened in the Q&A. It appears that he was not going to share this fairly important detail unless asked..

UMAPs have become the ultimate get out of jail free card if you want to say whatever fits your narrative. You want to claim no changes: overlay few colors, or way too many colors. You want to claim a change, circle a zone of the plot, scream that there is a difference. No one will bat an eye.

Did you know that you can use the #easystats #rstats {modelbased} package detect missed modeling by adding partial residuals to your plots? Check out the vignette on plotting options that shows you how to check your model: easystats.github.io/modelbased/a...

Let's Streisand-effect these people. Here's the editorial that somehow merited calling the cops. Read and share.

Did you recently update your #rstats to 4.6? Is your matrix algebra slow? Not #BLAS ting? `sessionInfo()` shows which BLAS library R uses. Switching from reference BLAS to Apple Accelerate is easily a 50x speedup. Let me know if this worked for you. 🚀 uatz.org/blog/2026-06-04-r-blas-macos/ #rstats

The mythology of academia - noble pursuit of knowledge for knowledge's sake etc - is used by administrators to convince academics who think they don't have a real job to do things against their own self interest. The sooner we realize we *are* workers - with employers - the better.

I uploaded a picture of a random root vegetable and asked ChatGPT to identify it. It told me it was a stochastic carrot.

SNL’s Tucker Carlson is so good, this has been my mental reaction these last weeks every time I see some non sense research practice

Also #singlecell RNaseq for sure 🧪🧬💻

Have a relaxing weekend, everyone who went to Cannes! #stats

Steve Quake quotes Brenner at #cshsymposium2026 : “biology is drowning in a sea of data and starving for knowledge”, arguing that never has this been more apt than now.

A survey of 2600 Ph.D. students, postdocs, and other researchers in 65 countries found that top three complaints were that supervisors: 1) used dismissive communication 2) provided little feedback 3) ignored team members’ well-being The main thing they wanted was empathy

When there’s a meeting and the first presenter is full python matplotlib, and the second presenter is full #rstats ggplot viz. I always feel like

🚀 Big updates are here! A new round of #rstats #easystats releases is now live on CRAN, featuring major performance boosts and new features for parameters, bayestestR, performance, modelbased, and see. Update via CRAN or grab the latest dev versions with: `easystats::install_latest()`

Same if you only use animal models but put the human disease term in the title

In a meta-analysis of 210 biomedical AI studies that statistically compared models under cross-validation, 97% used invalid statistical tests. Here's our new preprint doi.org/10.64898/202... led by @tianchu.bsky.social @hetuli.bsky.social @shaoshiz.bsky.social @nichols.bsky.social 1/N

"a paper today is a Frankensteined-together patchwork of supplementary files that run longer than the manuscript, datasets in repositories...code posted to GitHub without documentation. The paper has been replaced....by a mess" www.thetransmitter.org/from-bench-t...