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

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.

Thank you David! Officially a guiri!

Just reverted the video to explain protein folding!

Pinging @jingiyeo.bsky.social and @martinsteinegger.bsky.social

Very good point! It might worth investigating. We noticed this behaviour also when we clustered TED (over 81M singletons). that analysis was done at the domain-level, not at the chain level but the clustering wasn't that strict. Here I focussed more on the downstream from the domain end of things.

Amazing effort by @jingiyeo.bsky.social, @yewonhan.bsky.social, Andy Lau, @shaunkandathil.bsky.social, @hbkgenomics.bsky.social, Eli Levy Karin and @cathgene3d.bsky.social !

Explore AFESM with our website! You can search your favorite proteins from ESMatlas or AFDB using their identifiers. It's still a work in progress, with many exciting features on the way! Thanks @milot.bsky.social !

However, these novel domain combinations comprise only a small fraction (0.3%) of ESM-only clusters. The remainder are mostly low-quality predictions (53%), fragments (16%), known domains with potential unknown extensions (19%), or without identifiable domains (9.3%).

@yewonhan.bsky.social identified 11,941 novel multi-domain combinations!
We found membrane-associated domains (e.g., TonB dependent receptor), highlighting domain recombination rather than new folds as a driver of structural innovation in ESMatlas.

ESM-only clusters contain ZERO novel folds using the TED workflow. Re-modelling discarded domains (2.3M) with ColabFold revealed 1 novel fold; unlike AFDB’s >7k novel folds, hinting at a saturating fold space or ESMfold limitations.

With MGnify environmental labels, we computed the lowest common biomes per structural cluster, revealing protein adaptations unique to specific environments, especially extreme ones like hyperthermal, hypersaline, and glaciers.

We annotated ESMatlas with MMseqs2 taxonomy (93% coverage) and computed the lowest common ancestors (LCA). Most LCAs are at the superkingdom level, indicating structures shared across domains. Avg. clusters/genus: Bacteria 1,557; Archaea 723; Viruses 17; Eukaryotes 2 (sampling bias).