alphafold.bsky.social
Unofficial account exploring the intersection of biology, molecules, science, AI and protein folding with AlphaFold.
239 posts
9,511 followers
31,224 following
Prolific Poster
Conversation Starter
comment in response to
post
It’s AI Hype at its finest. It’s a slander and disrespect across so many disciplines of medical science. AI itself was already un-trustworthy, but now we see where the trained mistrust comes from.
comment in response to
post
Agreed.
comment in response to
post
more lines = more science
comment in response to
post
During the afternoon session we were joined online by Sergey Ovchinnikov who gave some invaluable insights into #alphafold
Sergey talked about AlphaFold3 and its limitations. If you are looking for alternatives, check out ColabFold ➡ bit.ly/3KQEHOe
@sokrypton.org
comment in response to
post
If you need help setting up, send me a message and I can send you the scripts I have for AlphaFold
comment in response to
post
it's great for stuff like, for example, AlphaFold, analyzing more 3D protein structures in an hour than a human could in a lifetime, and using that to tell us more about how proteins fold. Its not supposed to replace critical thought or user oversight or mechanistic understanding
comment in response to
post
Anna Kravchenko took participants through the HADDOCK3 tutorial for antibody-antigen modeling
This tutorial and other useful learnings including how to carry out antibody-antigen structure prediction from sequence using #alphafold can be found here ➡ bit.ly/45LCWgY
comment in response to
post
The thing that sucks so bad about this is that AI is going to truly revolutionize humanity, just no where near this fast or this extremely. AlphaFold is doing amazing things *now* and no one cares
They’re ruining an entire field for venture capital
comment in response to
post
It's getting there. Ignore GenAI slop, keep an eye on things like AlphaFold - that's where the real brilliance in this technology is: www.universiteitleiden.nl/en/news/2024...
comment in response to
post
There is more to AI than LLMs. There are real challenges that these systems and algorithms will solve long before AGI.
AlphaFold is a great example.
comment in response to
post
The result is that even competent AI researchers are totally off when they make predictions about what AI will be able to do for biology within the next 2-5 years. For example, they genuinely think AlphaFold solves all of protein biochemistry.
comment in response to
post
Instead, a different part of FANCM is required, the MM1 domain. We used AlphaFold modelling to show that the MM1 domain grabs onto the FA core complex binding and mutants in this domain also fail to repair DNA damage properly.
comment in response to
post
It'll be: THIS combination regulates THIS gene in THIS cell type... & so on. Will there be general principles? I wonder.
Another interesting aspect of this paper: the authors say AlphaFold-like algorithms aren't much help, for the simple reason that there is rather little data for training them./12
comment in response to
post
Using AlphaFold models from @bioinfo.se
and @pedrobeltrao.bsky.social’s work, we identified 5,800+ mutually exclusive protein interactions where two proteins have overlapping interfaces with another protein. This was greatly expanded during review using Jänes et al's resource (PMC11160786).