Compared to even a few months ago, it's surprisingly good. This was a single shot too
I like that it introduced a typo in John's message that Sarah doesn't see. That was a nice touch
AI introduces an update to the traditional OSI network model, whereby messages are wrapped in Enterprise-speak before being communicated to the network and decoded at the other end
I said Westdeutschen and not Westpreuรischen (I wanted to go back to 1983, not 1883)
Or maybe Claude's Bavarian
I'm very impressed with Claude right now
I transcribed a talk using whisper (local model) which completely messed up when I inserted a bit of German (as a joke)
When I had Claude "clean up the transcript," it recovered the original German almost perfectly
Teaser for presentation
Can't find the original PDF (chatGPT deep research is on it), but Margaret Dayhoff in 1967:
"There is a tremendous amount of information regarding evolutionary history and biochemical function implicit in each sequence and the number of known sequences is growing explosively"
Here is 2025 www.biorxiv.org/content/10.1...
Bioinformaticians have been writing the same intro for decades. Below is from 1983
Toward the end of the manual-sequencing era, the first generation of โautomated sequencersโ appeared. These machines did not initially automate much of the sequencing process. Gels were still prepared manually, and loaded manually. Only the readout of the sequence data was really automated and in the beginning the base-calling algorithms were quite unsatisfactory. Many steeped in the traditions of manual sequencing were doubtful that automation could compete with dedicated graduate student sequencers. The next few years would show how wrong we were.
I'm preparing this talk and going through some old papers to provide historical context
Bumped into this quote from doi.org/10.1093/nar/... which I think applies to many of today's AI skeptics (which I consider an increasingly-unsustainable view)
The early evidence that HIV causes AIDS was pretty iffy. Skepticism was not wrong
Eventually, the evidence became overwhelming
However, some never updated and became cranks
(This is a tweet about LLMs being stochastic parrots)
I will be speaking at BrisJAMS next week!
Like the โuneducatedโ explanation, the โreligiousโ one depends on a confusion between correlation and causation. Even if all women today with large families are religious, not all women who are religious have large families. Religion is correlated with total fertility, but not obviously causal.12 That is, something else mediates the association between religion and fertility. We donโt know what that is.
Some people confuse correlation and causation, some people confuse causation and โdeterministic direct 1:1 causationโ ๐ฅฒ
๐๐งฌ Our Environmental #Metagenomics course with @oskolkov.bsky.social & @luispedrocoelho.bsky.social is back (Oct 12โ16)!
Learn state-of-the-art metagenomic analysis with real Illumina & @nanoporetech.com data. From reads to high-quality #MAGs using modern #bioinformatics workflows
shorturl.at/Pi4Qt
"Planning to think about submitting sometime (Nature)"
Some links and tweets, first post of 2026!
luispedro.substack.com/p/links-twee...
New Year updates from the group with a focus on Anil Pokhrel who is working on metagenomics for food security!
bigdatabiology.substack.com/p/bdb-lab-ja...
There are some indices out there, whereby Australia has very high state capacity
๐๏ธโก If you use gzip/gunzip a lot in your pipelines, switch to the faster"libdeflate" versions instead! They use modern CPU capabilities to achieve a 2-3x speedup.
libdeflate is in conda, and "libdeflate-gzip" and "libdeflate-gunzip" are drop-in replacements. #unix
github.com/ebiggers/lib...
I woke up to some very sad news...
Peer was a giant of bioinformatics and computational biology and an inspiring mentor to everyone who had been in his group. I find myself repeating things he said all the time and will probably keep doing so for the rest of my life
www.embl.org/news/embl-an...
New substack: Some links & photos
Also, I'll be in Luxembourg through the end of the year and in Lisbon until the 10th of Jan. Get in touch if you want to meet up
luispedro.substack.com/p/links-phot...
I have a fondness for the pronouns meaning 'with X' that you find in Spanish, Portuguese and Galician (like conmigo 'with me' and contigo 'with you'), because both the con- bit and the -go bit come from the Latin for 'with' โ meaning that conmigo is etymologically 'with-me-with'.
Between these news and the astounding number of people that I know personally who had shingles even though they were "too young", I have decided to get the vaccine (and pay out of pocket) even though it is not part of the official recommendations for my age
"there's a new serif in town"
1. The European Parliamentโs attempt to ban plant-based foods from being sold as โsausagesโ, โburgersโ etc is a direct response to livestock industry lobbying. This is a short thread on how utterly bleeding ridiculous it is. ๐งต1/6
We're excited to release metaTraits.embl.de! ๐ฆ Interactively explore 140+ ๐บ๐ถ๐ฐ๐ฟ๐ผ๐ฏ๐ถ๐ฎ๐น ๐ฝ๐ต๐ฒ๐ป๐ผ๐๐๐ฝ๐ถ๐ฐ ๐๐ฟ๐ฎ๐ถ๐๐, harmonized & integrated from culture-derived collections ๐ฌ & genome-based predictions ๐งฌ for >2M MAGs & genomes.
Publication at NAR: doi.org/10.1093/nar/... @narjournal.bsky.social #microsky ๐งต 1/8
New version of macrel released (v1.6.0)
The biggest change is internal, using a much better approach to saving and loading models (thus removing the dependency on particular versions of scikit-learn)
A few other bugfixes were included
github.com/BigDataBiolo...
