Skip to content

AlphaLink2

Introduction

AlphaLink2 container is available on Euler in /cluster/apps/nss/alphafold/containers/AlphaLink2. This directory contains AL2.sif, the container itself, as well as the apptainer build file (alphalink2_recipe) and the AlphaLink2 weights, in a separate directory.

Run the container in an interactive job

The following instructions can also be adapted to run in a batch job.

First, request an interactive job. Feel free to adjust the resource requirements :

nmarounina@eu-login-44:~$ srun -n 4 -G 1 --time=04:00:00 --mem-per-cpu=40g --pty bash
srun: job 12215071 queued and waiting for resources
srun: job 12215071 has been allocated resources
nmarounina@eu-g1-030-2:~$ #here I am on a compute node on Euler

Here are the necessary binds. Please replace the location of your input files with the appropriate absolute path on the cluster :

nmarounina@eu-g1-030-2:~$ export APPTAINER_BIND=\
'/cluster/apps/nss/alphafold/containers/AlphaLink2/weights:/mnt/params:ro, \
/cluster/project/alphafold:/mnt/databases:ro, \
:/mnt/outputs, \
:/mnt/fasta,  \
:/mnt/crosslinks'

Then, get a shell inside of the container and activate the environment :

nmarounina@eu-g1-030-2:~$ singularity shell /cluster/apps/nss/alphafold/containers/AlphaLink2/AL2.sif
Apptainer>
Apptainer> . /opt/conda/etc/profile.d/conda.sh && conda activate alphalink
(alphalink) Apptainer> cd /opt/alphalink

At this point, you should be ready to use AlphaLink2. Here is an example of a command, taken from the AlphaLink2 Github and modified for the use on Euler. Please make sure to remove the backslashes and run it on a single line inside of the container :

bash run_alphalink.sh \
/mnt/fasta/<your_fastafile.fasta>\        # target fasta file
/mnt/crosslinks/<your_crosslinks.pkl.gz>\      # pickled and gzipped dictionary with crosslinks
/mnt/outputs \  #  directory for output files/folders
/mnt/params/AlphaLink-Multimer_SDA_v3.pt \    # model parameters, AlphaLink-Multimer_SDA_v2.pt is also available
/mnt/databases \    # directory of databases
2020-05-01 \                      # use templates before this date
20 \                              # use 20 recycling iterations (default: 20)
25 \                              # generate 25 sample (default: 25)
30 \                              # downsample MSAs to Neff 30 (default: -1, use full MSA, expects integer >= 1)
1                                 # integer > 0 activates this option. Remove MSA information for crosslinked residues (default: -1, use full MSA)