AlphaFold
AlphaFold is a software aiming to be able to predict 3D models of protein structures.
OnDemand
AlphaFold can be run in a browser as OnDemand service. This is the most straighforward and simplest way.
Kubernetes
Another option is to use Kubernetes service, which offers AlphaFold as an Jupyter notebook application.
Remote desktop
Using Remote desktop, AlphaFold can be also run from a Singularity image.
This option is recommended to advanced users or as a fallback option in case OnDemand and/or Kubernetes service are down.
This package provides an implementation of the inference pipeline of AlphaFold v2.1.1 This is a completely new model that was entered in CASP14 and published in Nature. For simplicity, we refer to this model as AlphaFold throughout the rest of this document.
Application is prepared as Singularity image. The image, together with databases and example scripts are available at
/storage/brno11-elixir/projects/alphafold
There are four models with two speed/quality tradeoff. All combinations of these parameters are prepared in example scripts in Metacentrum.
CUDA license
To use Alphafold with CUDA, you will also need to accept license for cuDNN library. You will find the link on MetaVO licence pages.
Models
You can control which AlphaFold model to run by adding the --model_preset= flag. We provide the following models:
monomer: This is the original model used at CASP14 with no ensembling.monomer_casp14: This is the original model used at CASP14 withnum_ensemble=8, matching our CASP14 configuration. This is largely provided for reproducibility as it is 8x more computationally expensive for limited accuracy gain (+0.1 average GDT gain on CASP14 domains).monomer_ptm: This is the original CASP14 model fine tuned with the pTM head, providing a pairwise confidence measure. It is slightly less accurate than the normal monomer model.multimer: This is the AlphaFold-Multimer model. To use this model, provide a multi-sequence FASTA file. In addition, the UniProt database should have been downloaded.
Speed/Quality
You can control MSA speed/quality tradeoff by adding --db_preset=reduced_dbs or --db_preset=full_dbs to the run command. We provide the following presets:
reduced_dbs: This preset is optimized for speed and lower hardware requirements. It runs with a reduced version of the BFD database. It requires 8 CPU cores (vCPUs), 8 GB of RAM, and 600 GB of disk space.full_dbs: This runs with all genetic databases used at CASP14.
Tips and useful information
Example of run with different models and speed/quality tradeoff with example file seq.fasta, multimer with multi.fasta. This test shows difference of RAM consuming a length of run with the same test.
| model | speed/quality | RAM | Duration | cluster - GPU |
|---|---|---|---|---|
| monomer | full_dbs | 185 GB | 28 min | glados - RTX2080 - 8GB |
| monomer | reduced_dbs | 153 GB | 36 min | glados - RTX2080 - 8GB |
| monomer_casp14 | full_dbs | 197 GB | 35 min | zia - A100 - 40GB |
| monomer_casp14 | reduced_dbs | 38 GB | 36 min | zia - A100 - 40GB |
| monomer_ptm | full_dbs | 190 GB | 36 min | gita - RTX2080 Ti - 11GB |
| monomer_ptm | reduced_dbs | 55 GB | 32 min | gita - RTX2080 Ti - 11GB |
| multimer | full_dbs | 119 GB | 75 min | zia - A100 - 40GB |
| multimer | reduced_dbs | 40 GB | 73 min | zia - A100 - 40GB |