Wolfgang Huber
@wkhuber.bsky.social
7.7K followers 2.3K following 510 posts
Statistician, Computational Biologist, R | Bioconductor. Love stats and ML. https://www.huber.embl.de Textbook: Modern Statistics for Modern Biology https://www.huber.embl.de/msmb/ (with @sherlockpholmes.bsky.social)
Posts Media Videos Starter Packs
Pinned
wkhuber.bsky.social
In July we had the 3rd edition of the Ukrainian Biological Data Science Summer School.
Now a regular, it has created a sense of community: young people eager to learn new topics, network across the country&world. Seeing how science is a positive&optimistic force.

www.huber.embl.de/group/posts/...
Ukrainian Biological Data Science Summer School 2025 – Huber Group @ EMBL
Uzhhorod, 19 July - 1 August 2025: the third edition
www.huber.embl.de
Reposted by Wolfgang Huber
erickratamero.com
Applications are open until 12/10 for both positions, so you might need to hurry :) I think these are both very exciting - and as a bonus (or a malus) your office space would probably be right next door to me!
Reposted by Wolfgang Huber
stollmeyer.bsky.social
🔥 Finally an EU leader says it out loud:

“We have been incredibly naive in entrusting our democratic space to social networks that are controlled either by large American entrepreneurs or large Chinese companies, whose interests are not at all the survival or proper functioning of our democracies.”
defenddemocracy.bsky.social
President Macron: “Europeans, let's wake up!

We have been incredibly naive in entrusting our democratic space to social networks.”

defenddemocracy.eu/macron-democ...
Reposted by Wolfgang Huber
gautamdey.bsky.social
If you’re applying to your dream lab for an internship/PhD/postdoc, always send a second email 1-2 weeks after the first one if you don’t hear back.

I promise we will be grateful rather than annoyed. My email inbox is a disaster and I’m quite junior - and very few of us have secretarial support
wkhuber.bsky.social
(Huber lab)[EMBL] welcomes applications for a PhD in ML for spatial omics [representation learning + integration with biostatistics, cell + anatomy foundation models, collaborate with domain scientists on cancer and dev:bio discovery science] |> Apply though the ELLIS portal ::: Deadline 2025-10-31
Reposted by Wolfgang Huber
p-hunermund.com
Post a favorite econ paper: Lichter et al. (2021, JEEA) @sigginho.bsky.social

Regions with higher Stasi spy density in East Germany still show lower trust, lower incomes, higher unemployment, and less entrepreneurship decades after reunification.

Open-access link: doi.org/10.1093/jeea...
Reposted by Wolfgang Huber
Reposted by Wolfgang Huber
burritojustice.bsky.social
quick while the government is shut down let’s all switch to metric
wkhuber.bsky.social
Exactly! Now that does sound a lot like image segmentation, which can be formulated as simultaneously minimising distances (a) among pixels and (b) among their values (intensities).

Real segmentation algorithms can even do more fancy things, like take into account texture, or use boundary markers.
wkhuber.bsky.social
but the commonalities seem big enough to not justify a completely different name and different set of computational methods?
wkhuber.bsky.social
to a metric space (tuples of color intensities or expression levels), such as that distances with a segment (or cluster) in both spaces are simultaneously mimised. I understand that there will be different trade-offs, different useful heuristics etc. depending on application domain and objective,...
wkhuber.bsky.social
Thank you! I agree, but I'd call these differences "marketing"; afaIcs, the underlying maths is the same, and algorithms should to a good extent by transferable?

The common mathematical task being something like: partition a metric manifold (or a lattice graph, etc), based on a function from it...
wkhuber.bsky.social
Yes, the remark was tongue-in-cheek of course.
There will always be plenty of work for statisticians to deal with observational data, or experimental data from non-ideal assays and designs.
wkhuber.bsky.social
Skymind: I'm learning about 'spatially aware clustering' in spatial omics and wondering how it is different from 'segmentation'? (in essence, not by way of marketing)
wkhuber.bsky.social
In other words: a good statistician makes themself superfluous.
Reposted by Wolfgang Huber
ewanbirney.bsky.social
Where most of the drama is in most experiments. Classic if you needed fancy statistical test, you didn't design your experiment well enough :)

Though if anyone has a power/design framework for AI interpretation please let me know. It's the wild west out there!
wkhuber.bsky.social
Am contributing to a collaborative dry-wet project proposal and everyone incl. myself thinks I should be writing about fancy algorithms. Instead, I spend all time on experimental design,power, what do we know or not, what do we aim to show? And on 2nd thought, that's probably OK for a statistician.
Reposted by Wolfgang Huber
dhrebik.bsky.social
This comic is becoming increasingly realistic.
m.xkcd.com/3056/
Reposted by Wolfgang Huber
floravincent.bsky.social
Dear all, registration for the upcoming EMBO/EMBL Symposium on "the cellular mechanics of symbiosis" is now OPEN 🤩 ! Feel free to register ! www.embl.org/about/info/c...
#EESSymbiosis #ProtistsOnSky #SymbioSky
@hassansalem.bsky.social @berasymbionts.bsky.social @embl.org
The cellular mechanics of symbiosis: sensing friend from foe
www.embl.org
wkhuber.bsky.social
A useful analogy.

One can say that much of the progress in computer programming is aligned with ever increasing levels abstraction.
LLM-based coding tools are "just" a next level.

The choice of objective, and the measure of success, still remains something external to the tool.
albertvilella.bsky.social
The new trend of #AI #LLM tools that help people achieve their tasks faster isn't too different to the shift in Computer Programming between the pre-compilers world and the world were compilers allowed scientists to define computer languages in a more natural way:
wkhuber.bsky.social
Do you like theoretical biology, mathematical&conceptual modelling and want to apply yourself to cutting-edge research at all levels of biology, from molecules to ecosystems?

Come to EMBL! Scope includes physics-based models, emergence, statistics, ML/AI
embl.wd103.myworkdayjobs.com/en-US/EMBL/j...
Reposted by Wolfgang Huber
odedrechavi.bsky.social
How every PhD project develops