Jameel Abduljalil
@jameel-abduljalil.bsky.social
PhD candidate in (Computational) Structural Biology at the University of Sydney.
Personal account.
Personal account.
Clorox was suggested as an antiviral drug a few years ago.
May 4, 2025 at 9:43 AM
Clorox was suggested as an antiviral drug a few years ago.
Maybe the max. got mistakenly divided by 10.
April 21, 2025 at 10:57 AM
Maybe the max. got mistakenly divided by 10.
Exactly! Structural diversity (not just data quantity) is the bottleneck. My original point was that even retraining on the entire PDB content is expected not to significantly fix the limitations. Some plant proteins still have no homologs to build MSAs let alone predict their 3D with confidence.
February 12, 2025 at 12:34 AM
Exactly! Structural diversity (not just data quantity) is the bottleneck. My original point was that even retraining on the entire PDB content is expected not to significantly fix the limitations. Some plant proteins still have no homologs to build MSAs let alone predict their 3D with confidence.
AF3 limitations are likely due to its training data. Even with retraining on the entire PDB, some biases and inaccuracies are expected to persist. This is because the PDB itself is skewed towards structures that are easier to determine experimentally and numerous entries have similar folds.
February 9, 2025 at 7:08 AM
AF3 limitations are likely due to its training data. Even with retraining on the entire PDB, some biases and inaccuracies are expected to persist. This is because the PDB itself is skewed towards structures that are easier to determine experimentally and numerous entries have similar folds.
It would be 1:00 AM here 😀
Will watch the YouTube recording later on.
Will watch the YouTube recording later on.
January 22, 2025 at 10:20 AM
It would be 1:00 AM here 😀
Will watch the YouTube recording later on.
Will watch the YouTube recording later on.