📢 Rfam 15.1 is here!
✨ 50 new RNA families including riboswitch candidates, plastid ncRNAs, snoRNAs, plant xrRNAs and more.
🧬 10 families updated with 3D structures
🖥️ Brand new interactive alignment viewer
Take a look xfam.wordpress.com/2026/01/08/r...
#RNA #Bioinformatics #RNAbiology
🎉 RNAcentral Release 26 is here! This release introduces our biggest structural change yet: gene-level entries for ncRNAs across 204 organisms.
For the first time, you can explore RNA data at the gene level, not just individual sequences.
🧵👇
I'll be presenting the @rnacentral.bsky.social & @rfamdb.bsky.social poster "Integrating the RNA Universe: Advances and Future Directions in RNAcentral and Rfam Resources" today at #ismbeccb2025! Come and say hi at board C-235 if you'd like to chat about our current status and future plans! 🧪
One of my favorite talks from today! But I do love a graph
Course at EMBL-EBI Data science for life scientists 16 - 21 June 2025 Hinxton, United Kingdom
If you're a #lifescientist looking to develop your skills in #datascience, including using @python.org and the applications of #AI and #machinelearning then this course is for you!
Applications are closing soon - you have until 2 March: www.ebi.ac.uk/training/eve...
🧬🖥️🧪 #GeneSky
Hi Erik, I'm a research Fellow at EMBL-EBI working on AI curation for non-coding RNA: orcid.org/0000-0002-82... Would you be able to add me to the science feed please? Thanks!
This is my first first-author publication while at @ebi.embl.org! Its a good one, using LLMs to do some literature curation in non-coding RNAs. We've got big plans to do even more cool stuff with LLMs in the near future!
Gratifying that I'm still better at prompt engineering (for given values of better) than Claude
First impressions of the 2nd section test piece for the brass band championship regionals this year: its a banger! That last section is great!
I'm sure 3 more months of rehearsal on it will knock that attitude out of me though
Huh, interesting that bsky stripped my 34 spaces.... Anyone interested will have to provide their own
Just typing a bunch of spaces into that app, you see it cycle through up to 81 different tokens to represent just a load of spaces (I think I know where that magic number comes from though). It's wild that 81 spaces being a single token is coming out of the tokenizer training though
Fiddling about with LLMs and their weirdness... Why do loads of tokenizers have one single token for 34 consecutive spaces?
Go here: huggingface.co/spaces/Xenov... and try pasting in " " (without the "). It's token ID 9898 for GPT4 and Llama3
