|
- People are overestimating Alphafold and it’s a problem
Alphafold is an amazing tool, but please please please, don’t think it’s omniscient and just magically knows what proteins look like Especially considering Proteins with IDRs or ones with a Lack of similar templates
- alphafold - Reddit
r alphafold: A subreddit to discuss AlphaFold and protein peptide research
- Is Folding At Home Still Relevant And Worth The Effort? - Reddit
Folding@home uses mechanistic modelling, allowing the time-related process to be understood more AlphaFold uses machine learning, which although has impressive final-state accuracy, doesn't currently give any data between start and final states But I'm not sure if this is more to do with lack of 'training' data for the ML?
- How to do useful stuff with Alphafold : r bioinformatics - Reddit
Alphafold is all nice and good and I was able to display the proteins as a 3D model in blender But how do you do really useful stuff like actually working with the structure in the raw data form (atom distance matrix)? I want to do phylogenetic comparisons with other proteins and basically analyze the substructures regarding electric charges and stuff like that I know this can be programmed
- AlphaFold 3 is a Fantastic Breakthrough and Deserves all the . . . - Reddit
AlphaFold is helping researchers build models, but it can’t yet simply find all protein states F@H has always focused on molecular dynamics, while Rosetta@Home (a BOINC project) looked for protein structures and has been more supplanted by AlphaFold F@H also open sources all its results and was probably used to train AlphaFold
- Running AlphaFold on a standalone PC? : r bioinformatics - Reddit
The AlphaFold GitHub pages only discuss executing the program in Google Cloud, and the specs for their Google Cloud instance sound quite formidable: "a [virtual?] machine using the nvidia-gpu-cloud-image with 12 vCPUs [virtual CPUs?], 85 GB of RAM, a 100 GB boot disk, the databases on an additional 3 TB disk, and an A100 GPU "
- How important of a breakthrough is DeepMinds AlphaFold? - Reddit
The problem with AlphaFold is it's trained on a limited set of known data, specifically annotated data where a large number of similar structures exist It sucks at predicting membrane proteins and some barrel proteins I fed it, as the active sites tend to be cocked up to the point where it can't fit known substrates
- Announcing AlphaFold 3: our state-of-the-art AI model for . . . - Reddit
AlphaFold's ability to rapidly and accurately predict protein structures has accelerated research across various medical fields, enabling deeper understanding of disease mechanisms and facilitating drug discovery efforts [1] [3] Citations:
|
|
|