A dog laughing while sitting in the park.

Obligatory national pet day post. She laughs at all my jokes despite the overwhelming evidence that she shouldn't.

1/ Hot take: deep learning is overkill for 90% of bioinformatics problems.

I've watched teams spend months building neural networks that get outperformed by XGBoost trained in an afternoon.

Here's why simpler models keep winning.

Bluesky Map Interactive map of 3.4 million Bluesky users, visualised by their follower pattern.

I made a map of 3.4 million Bluesky users - see if you can find yourself!

bluesky-map.theo.io

I've seen some similar projects, but IMO this seems to better capture some of the fine-grained detail

there’s nothing better than reaching into a bag of pistachios and pulling out one that’s already open...

...there’s also nothing worse than reaching into that same bag and pulling out just an empty shell.

Mars Express image of Olympus Mons volcano showing frost, 2024

In the difficult years ahead, we should remember that the snows of Olympus lie silent beneath the stars, waiting for our grandchildren.

- Arthur C. Clarke

Happy new year nerds! Episode 2 is out! Go listen to my conversation with Dr. Swetha Murthy and learn how she found her confidence through her science 🧬 Available now everywhere you listen to podcasts!

Austin Schubert (Saunders Lab) won "Most Innovative Approach" at the Cure SYNGAP1 conference. His poster was titled "Brain–wide discovery of when, where, and how SYNGAP1 haploinsufficiency alters neurodevelopment through single cell RNA profiling.”
cc: @trill-nye.bsky.social @your-arpy.bsky.social

Full house for our Bi[o]hub workshop on Day 3!
Communities aren’t just built, they are co-created.
Together, we’re shaping an open, connected world of all-things-single-cell research, where sharing tools and ideas drives discovery.

Screenshot of blog post at https://dirk.eddelbuettel.com/blog/2025/10/26#duckdb-mlpack_0.0.2 announcing that mlpack is now a duckdb community extension

duckdb-mlpack 0.0.2: mlpack is now a duckdb community extension
Bringing mlpack machine learning to duckdb SQL
dirk.eddelbuettel.com/blog/2025/10...

Human brain cell types shape host-rabies virus transcriptional interactions revealing a preexisting pro-viral astrocyte subpopulation How virus-host cell interactions and innate immune antagonism shape neurotropic infection dynamics across diverse brain cell types is largely unknown.…

Huge congratulations to my awesome lab mates, @your-arpy.bsky.social, and their collaborators @pasteur.fr on publishing their new paper showing how rabies hijacks brain cells and revealing a rare astrocyte subtype that boosts viral transcription 🧪 #SingleCell

www.sciencedirect.com/science/arti...

Probably just the duck from DuckDb running a quick query. It'll be gone soon.

Happy to share my first publication! AnnSQL leverages DuckDB to create a relational database based on each AnnData layer, enabling fast queries on massive, atlas-sized single-cell datasets. 🧬🖥️

#Bioinformatics #SingleCell #DuckDb

GitHub - ArpiarSaundersLab/annsql: The AnnSQL package enables SQL based queries on AnnData objects. The AnnSQL package enables SQL based queries on AnnData objects. - ArpiarSaundersLab/annsql

We're excited to offer the first #SQL based single-cell exploration and analysis tool. If you encounter anomalies or bugs, please report to the AnnSQL GitHub issue board: github.com/ArpiarSaunde...

𝗦𝗽𝗲𝗲𝗱: AnnSQL was built for querying large datasets.

tl;dr; If your dataset is > ~100k cells or out of memory, consider using AnnSQL. If your dataset is < ~100k cells or within memory, Scanpy or Seurat are the winners.

𝗜𝗻𝘁𝗲𝗿𝗼𝗽𝗲𝗿𝗮𝗯𝗶𝗹𝗶𝘁𝘆: AnnSQL databases (.asql) can be shared and also play nicely with a variety of languages; even R!

docs.annsql.com/interoperabi...

𝗠𝗲𝗺𝗼𝗿𝘆: Using the #DuckDb in-process database engine behind the scenes, AnnSQL is memory respectful. AnnSQL also enabled a chunking parameter for all preprocessing methods— making it a powerful ally capable of processing massive datasets on a laptop.

𝗗𝗼𝗰𝘂𝗺𝗲𝗻𝘁𝗮𝘁𝗶𝗼𝗻: Tutorials and API usage can be viewed at: docs.annsql.com

𝗡𝗲𝘄 𝗙𝗲𝗮𝘁𝘂𝗿𝗲𝘀:
- Save raw layer
- Filter by cell & gene counts
- Highly variable gene selection
- PCA (experimental)
- Leiden clustering, UMAP, Diff Expression
- Various plotting utilities

𝗣𝗿𝗲𝗽𝗿𝗶𝗻𝘁: In our updated preprint, we runtime profile filters vs queries using AnnSQL, AnnData, and Seurat.

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

GitHub - ArpiarSaundersLab/annsql: The AnnSQL package enables SQL based queries on AnnData objects. The AnnSQL package enables SQL based queries on AnnData objects. - ArpiarSaundersLab/annsql

𝗔𝗻𝗻𝗦𝗤𝗟 v1 has landed! Explore atlas sized single-cell datasets, with minimal resources using SQL that's powered by the #DuckDb database engine. #SingleCell 🧬 🖥️ 🖖

github.com/ArpiarSaunde...

Happy to see lots of DuckDuckGoers. It's also worth considering Brave search. 🦁

search.brave.com

Whoa, that's dedication 💪 Hope you find what you're looking for soon!

Currently using ScEasy or SeuratDisk. Both are great tools and I've had some success using both, however, they sometimes can fail depending on the dataset and the layers present. The interoperability page you linked has some options I've yet to explore. Thank you for sharing!

I spend way too much time converting AnnData objects into Seurat objects and haven't found a tool or pipeline that just works, while retaining all data layers. What are some tools or pipelines to ease this process? 🧬 🖥️ #SingleCell

By extension, Wikipedia says DeepSeek trained on its data and asks for $5 donation.

Learning about some of the magic behind the #DeepSeek R1 paper: "Optimisations actually began at chip level, below even CUDA, with DeepSeek engineers writing assembly language to overcome H800 bandwidth throttling. This in itself is an extraordinary technical feat"

AnnSQL: A Python SQL-based package for large-scale single-cell genomics analysis on a laptop As single-cell genomics technologies continue to accelerate biological discovery, software tools that use elegant syntax and minimal computational resources to analyze atlas-scale datasets are increas...

SQL is way underappreciated in #bioinfo. Our lab just dropped a preprint for a tool (AnnSQL) that uses DuckDb for single-cell analysis.

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

Usually I try to reduce it to an easily digestible analogy. Something like... Biology is like a giant puzzle and we use computers to try and put all the pieces together.

“It’s easier to fool a man than to convince him that he’s been fooled.” - Mark Twain

Well, someone tried to tell me that Mark Twain never actually said that, but I don’t believe them

Commander Data from Star Trek The Next Generation as a change my mind meme stating "AI is just matrix multiplication. Change my mind."

The numbers speak for themselves.

#startrek #startrekmemes #data #ML #AI 🖖