Pierre-Paul Axisa

SAFE-LD: A novel method for the estimation of linkage disequilibrium from summary statistics Genome-wide association studies (GWAS) have greatly advanced our understanding of the genetic architecture of complex traits. Downstream analyses of GWAS summary statistics require accurate in-sample ...

I am very happy to finally share SAFE-LD, a method convert genotypes to an anonymised privacy-preserving version to be used to precisely compute LD. Led by @gdesanctis.bsky.social Claudia Gianbartolomei @sodbo.bsky.social and Davide bolognini.

📄: shorturl.at/BgXqq
Github: shorturl.at/b49Wa

With so many new methods buried in results-focused papers, this thread is my encouragement to groups to also publish separate protocols papers! You'll be pleasantly surprised at the engagement these papers get, and most importantly they provide real impact aiding others in their science.

AutoSpectral improves spectral flow cytometry accuracy through optimised spectral unmixing and autofluorescence-matching at the cellular level The advent of spectral flow cytometry has seen a rapid rise in the complexity of flow cytometry experiments, allowing the construction of assays with at least 50 fluorescent parameters. To correctly d...

A small 🧵 encouraging people to publish protocols.

Our #AutoSpectral release has made me reflect on how much the community tends to appreciate protocol papers. People rarely present protocols as their main achievements, but they are often the most impactful work.

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

To be clear, I’m not blaming people for not maintaining their website, most academics get no credits from funders and their institutions to maintain these essential resources. But there are mitigations to safe guards the data for the community

PSA: it’s all great if you provide an interactive dashboard to browse the data you painstakingly generated. But please do back up the data on persistent archive like zenodo. It’s too bad we can’t access the database when inevitably the server goes down and you moved onto other projects/institutions

🧬 EuroBioC2026 is coming to Turku, Finland, from June 3–5, 2026!

Expect 3 exciting days with the Bioconductor community, showcasing the latest advances in Bioconductor software and emerging technologies shaping computational biology.

👉 More details soon at eurobioc2026.bioconductor.org

SARS-CoV-2 mRNA vaccines sensitize tumours to immune checkpoint blockade - Nature mRNA vaccines targeting SARS-CoV-2 also sensitize tumours to immune checkpoint inhibitors.

I've been asked about this paper (www.nature.com/articles/s41...) enough times that I finally had to read it.

The basic claim is that mRNA vaccines had huge synergistic benefit with checkpoint blockade against lung cancer melanoma.

tl;dr it's glam-slop & everyone involved should be embarrassed

Police sources actually say the robbers were postdoc trying to raise funds for prelim data to go in their faculty applications

Yes! It’s actually that simple 😅. Code completion is so good on VS code

That can do, I was just wondering how I could get this behavior when with a keyboard shortcuts which would reformat the path string when I select it.

I should have said, I work mostly with VS code, but some Rstudio too

a man in a suit and tie is sitting on a bench and making a funny face . ALT: a man in a suit and tie is sitting on a bench and making a funny face .

#rstats : when you have a path string like "this/is/my/path", is there a formatting shortcut somewhere to quickly split it into a vector like c("this", "is", "my", "path") (even better if with {here} or {this.path})

{lintr} is harassing me for long lines and I ought to do something

That’s good, chat-GPT is an academic! If you want to claim you’re the first to solve it, just "forget" to cite previous contributions 😜

Be careful, check your kids’ book, they might contain rogue pseudotime analyses 💻🧬

Qualpal: Qualitative Color Palettes for Everyone Larsson, J., (2025). Qualpal: Qualitative Color Palettes for Everyone. Journal of Open Source Software, 10(114), 8936, https://doi.org/10.21105/joss.08936

💡 Came across this nice tool today:

🎨 qualpal for algorithmically choosing maximally distinct colors under certain restrictions #dataviz

JOSS paper, online tool, R package #rstats

joss.theoj.org/papers/10.21...

Oh no, working during lunch is very much frowned upon! I think that’s a good rule tbh

You should try France, half hour does not exist. It’s minimum an hour, and a Russian roulette on whether you’ll wrap up at the one hour mark or end up in a 3 hours meeting 🤪 🤯🫠😭🤷‍♂️😵

Aka parametric parking

Are you going to #ASHG2025 next week in Boston? Interested in getting some mentorship from leading figures in the field (and me)?

Check out our speed mentoring event on Thursday! Sign-ups are online so reserve a spot ASAP.

site.pheedloop.com/event/ASHG25...

Eh le Monde, il faudrait faire quelques révisions d’anatomie! "thymus, organe situé dans la partie supérieure gauche de l’abdomen près de l’estomac."

RE contacting committees; even if I did not make the cut, contacting people at the institute helped narrow down how I should shape my proposal. Things not explicit in the call, eg wet/dry lab balance. They might be already "staffed" on some niches so you can adapt your materials to that.

a group of men are sitting at a table and one of them is saying you gotta believe ALT: a group of men are sitting at a table and one of them is saying you gotta believe

From having been thru this recently, I also feel like there’s just a sheer amount of luck, so you need to put out a large volume of applications and hope some of them go thru… That’s what Friends/colleagues said when I started to waver:

It is really hard to get any feedback when you don’t make it past the first stage, which is really frustrating… I guess one advice is to try to get colleagues who have xp being on committees review your materials beforehand. (But it is time consuming for them)

💯. I would add *a manifest* that just counts how many samples of which type you have. So trivial! But sometimes you are missing samples. Or you think you filtered out x samples but did not. Printing before/after data inventory can save many headaches!

Cartoon by the excellent Allison Horst (https://allisonhorst.com/git-github). Includes a quote from Hadley Wickham and Jenny Bryan that says “Using a Git commit is like using anchors and other protection when climbing…if you make a mistake you can’t fall past the previous commit. Commits are also helpful to others, because they show your journey, not just the destination.” On the right, two little monsters climb a cliff face. Their ropes are secured by several anchors, each labeled “Commit”. Three monsters on the ground support the climbers.

One of the biggest increases in my analysis/coding productivity in recent years is completely embracing git repos 🐧

Coupled with GitHub it is *so easy* to be open, reproducible, efficient, and accountable to myself and colleagues 👨‍💻

No more chaotic servers wondering which script led to what! 😵‍💫

Disappointment Disappointed GIF ALT: Disappointment Disappointed GIF

Is it me or now Rstudio saves files with .r instead of .R? Same goes for Rmd vs rmd. What is this heresy #rstats? Well I actually don’t know if there is one way that is more correct but

I thought (naively) that an LLM needs massive amounts of data for training. Would this be pretrained LLM + fine tuning with your own data? Or is it that with recent advances it would need less training data, so could be only your own?