iiinova — where your digital self keeps evolving

ProFam: Open-Source Protein Family Language Modelling for Fitness Prediction and Design

To advance the family-based modelling approach, we are releasing the entire framework open source: ProFam Atlas: A curated, large-scale training corpus containing nearly 40 million protein families. Code & Weights: github.com/alex-hh/prof... Data: zenodo.org/records/1771...

biorxiv.org

Protein language models have become essential tools for engineering novel functional proteins. The emerging paradigm of family-based language models makes use of homologous sequences to steer protein ...

031

For design, ProFam-1 excels at homology-guided generation. It produces diverse sequences with low sequence identity to natural proteins while preserving predicted structural similarity and conservation patterns of the natural family, even when conditioning on just a single example sequence.

121

By conditioning on homologous sequences, ProFam-1 is competitive with state-of-the-art zero-shot fitness prediction on ProteinGym, outcompeting much larger PLMs such as ESM.

120

judewells.bsky.social @judewells.bsky.social · 18/09/2025

Built by CATH, TÜM and NVIDIA, ProFam-1 is our new open-source protein family language model (pfLM) designed to generate functional protein variants and predict fitness using in-context example sequences.

1115

Reposted by CATH-Gene3D

It was lovely to speak at the CATH 30 symposium, celebrating 30 years of the @cathgene3d.bsky.social protein structure classification database. I was presenting recent work on our new generative protein-family language model: preprint coming soon.

0113

Rob Finn on MGnify, everything bacteria and functions in different environments

021

Now Maria Martín from UniProt is telling us how AI-based tools are shaping the future of one of the key resources for protein sequences and function.

121

From structures to sequences, now Alex Bateman and the quest to annotate and classify all proteins!

121

Starting our afternoon session with a talk by Sameer Velankar, of PDBe and AFDB fame among other endeavours!

And now @gonzaparra.bsky.social on his first talk on protein frustration as a PI! Well done!

162

David Jones, on novel folds in AFDB and CATH’s founding being celebrated at a now-closed Pizza place in Euston Station

From CATH to Computational Enzymology, Dame Janet Thornton on the birth of CATH and beyond!

First Keynote by Burkhard Rost, on the impact of protein language models on the field of structural biology

Kickstarting our symposium “Protein Annotations in the age of AI” at UCL!

172

If you'd like to showcase your research with a poster, details are included in the registration page. We hope to see you there!

001

We have a stellar lineup of speakers! Christine Orengo Burkhard Rost Janet Thornton David Jones Gonzalo Parra @gonzaparra.bsky.social Sameer Velankar Alex Bateman Maria Martin Rob Finn Gerardo Tauriello Alexey Murzin

132

There will be talks from world leaders in structural bioinfomatics on various themes including pioneering protein language models and key international resources including: PDBe, InterPro, UniProt, MGnify, SWISS-MODEL, FrustraEvo and CATH.

112

Protein Annotations in the age of AI

CATH turns 30 years old this year! We are organising a 1-day symposium on September 16th at UCL, highlighting recent AI-based developments to enhance protein family classifications, annotations and analyses. www.eventbrite.co.uk/e/protein-an...

eventbrite.co.uk

A not-for-profit symposium hosted at UCL - more details about speakers and venue below.

2127

CATH-Gene3D @cathgene3d.bsky.social · 20/08/2025

Another CATH outing at Greenwich Park after a lovely cruise along the Thames and a pub lunch!

010

Reposted by CATH-Gene3D

Nicola Bordin @nbordin.bsky.social · 28/04/2025

Our latest preprint is out on bioRxiv! A collaboration between the groups of @martinsteinegger.bsky.social , David Jones and Christine Orengo, we clustered AlphaFold Database and ESMatlas, a whopping 821 million proteins! We reveal biome-specific groups & over 11k novel domain combinations.

biorxiv.org

Metagenomic-scale analysis of the predicted protein structure universe

Protein structure prediction breakthroughs, notably AlphaFold2 and ESMfold, have led to an unprecedented influx of computationally derived structures. The AlphaFold Protein Structure Database now prov...

23913

Reposted by CATH-Gene3D

EMBL-EBI @ebi.embl.org · 03/03/2025

TED is a collaborative project between the structural bioinformatics groups of Professor David Jones & Professor Christine Orengo @cathgene3d.bsky.social at @ucl.ac.uk. The TED integration is set to enhance the interpretability and usability of #AlphaFold predictions. Is this useful in your work?

0102

Reposted by CATH-Gene3D

EMBL-EBI @ebi.embl.org · 03/03/2025

🚀 #AlphaFold Database update AlphaFold DB now integrates The Encyclopedia of Domains (TED) – a resource designed to systematically identify & classify structural domains within AlphaFold-predicted protein structures. www.ebi.ac.uk/about/news/u... @pdbeurope.bsky.social

111844

We have updated sequences in our Functional Families by scanning FunFam-HMMs against UniProt release 2024_02, giving a 276% increase in FunFams coverage. The mapping of TED structural domains has resulted in a 4-fold increase in FunFams with structural information.

000

New PDB and TED data increases the number of superfamilies from 5841 to 6573, folds from 1349 to 2078 and architectures from 41 to 77.

100

CATH v4.4 represents an expansion of ∼64 844 experimentally determined domain structures from PDB. We also present a mapping of ∼90 million predicted domains from TED to CATH superfamilies.

100

We report a significant expansion of structural information (180-fold) for CATH superfamilies through classification of PDB domains and predicted domain structures from the Encyclopedia of Domains (TED) resource.

100

For those without access to the Science article, we added a full access link on the TED website (ted.cathdb.info) landing page! 6/6

000

We will keep TED up-to-date with further releases of the AlphaFold Protein Structure Database and we hope TED will help your future research! 5/6 🧵

110

All data and code to reproduce TED and TED-web are available in Zenodo zenodo.org/records/1390... Here you can find per-chain and per-domain annotations, domain interaction data and model weights 4/6 🧵⤵️

100

Ian Sillitoe made a beautiful web interface for TED (ted.cathdb.info), which allows the community to look up their favourite UniProt Identifier, visualise domain annotations and associated metadata on quality and interactions. 3/6 🧵⤵️

100

Compared to our bioRxiv version, we processed TED redundant in the same fashion as TED100, swapping the previous multi-chain approach with a single-chain approach. Each of the 214M chains in TED is now being treated equally, identifying 365 million protein domains. 2/6 🧵⤵️

100