I’m late to my own party but excited to share our new work on predicting SLiM-mediated protein-protein interactions, out today in @naturemethods with Joe Cunningham, @GregKoytiger, and @sorger_peter! A blogpost is forthcoming but for now a tweetstorm (1/8)https://www.nature.com/articles/s41592-019-0687-1 …
-
Prikaži ovu nit
-
The paper’s key technical innovation is the formulation of an ML-derived energy model that shares energy potentials between protein families, enabling information transfer from data rich domains (e.g. SH2s) to relatively data poor ones (e.g. PTBs) (2/8)pic.twitter.com/MmnZ9O9hp8
1 reply 1 proslijeđeni tweet 7 korisnika označava da im se sviđaPrikaži ovu nit -
This gets us state-of-the-art performance for domains that weren’t well-characterized previously, which is perhaps unsurprising, but also SOTA results for ones that have been studied to death (ahem, SH2s, ahem) (3/8)pic.twitter.com/dO8UEeKjdy
1 reply 0 proslijeđenih tweetova 4 korisnika označavaju da im se sviđaPrikaži ovu nit -
The model is mechanistic and interpretable in the sense that we can decompose a multidentate protein-protein interaction, comprised of multiple peptide-binding domains and SLiMs, in terms of the underlying interactions between them (4/8)pic.twitter.com/O7e8y2bMDQ
1 reply 0 proslijeđenih tweetova 8 korisnika označava da im se sviđaPrikaži ovu nit -
It’s also mechanistic in the sense of providing residue-level energetic and structural insights consistent with our understanding of the domain families we have modeled (SH2s, SH3s, PDZs, WWs, PTBs, WH1s, and tyrosine kinases and phosphatases) (5/8)pic.twitter.com/nwu99VyRaW
1 reply 0 proslijeđenih tweetova 6 korisnika označava da im se sviđaPrikaži ovu nit -
For now we used it to do a reasonably standard network-level analysis of the human proteome, which has revealed interesting aspects of the topological organization of peptide-binding domains. Stay tuned on this front however as we’ll have a lot more to say very soon (6/8)pic.twitter.com/2bqjykWrzW
1 reply 0 proslijeđenih tweetova 8 korisnika označava da im se sviđaPrikaži ovu nit -
The code is on GitHub (https://github.com/aqlaboratory/hsm …), and there’s a website for interactive predictions and visualizations as well as bulk downloads (http://proteinpeptide.io ) (7/8)
1 reply 2 proslijeđena tweeta 9 korisnika označava da im se sviđaPrikaži ovu nit -
Finally what makes this particularly special for me is that it’s Joe’s first paper! Joe came to work with me and Peter Sorger as an MIT undergraduate and it has been a singular joy to work with him throughout this time. His only vice is that he refuses to get on Twitter ;-) (8/8)
1 reply 0 proslijeđenih tweetova 17 korisnika označava da im se sviđaPrikaži ovu nit
Here is the promised blogpost: https://protocolsmethods.springernature.com/users/346093-mohammed-alquraishi/posts/58576-imbuing-machine-learning-with-a-hint-of-biophysics … on the motivation behind our modeling / ML approach for protein-peptide interactions. We'll likely have another post soon focused more on the (structural) biology.
Čini se da učitavanje traje već neko vrijeme.
Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.