Snakemake workflow: `AF_server`

A Snakemake workflow to serve AlphaFold queries through an email server

Setup

Install Snakemake, set up AlphaFold
Clone this repository
Edit the configuration file config/config.yaml (see README inside the config/ directory)
If on HPC, edit config/envmodules.yaml to load necessary modules through module load
Submitting queries must be explicitly allowed by editing the whitelist config/whitelist and adding allowed source email addresses
SLURM configuration is handled by the slurm/ snakemake profile

Usage

To invoke the pipeline, run:

# password to the inbox/outbox
export EMAIL_PASS=...
snakemake -j 8 --rerun-incomplete --cores 16 --profile slurm --use-conda --use-envmodules

The server can be invoked automatically in a crontab job every few minutes.

The password to the email account used to serve the queries must be passed through the environmental variable EMAIL_PASS

The --j flag specifies the number of parallel snakemake instances (i.e. maximum numbers of target that can be modelled at the same time)

The --use-conda flag allows to setup the environment for AlphaFold (tested on v2.2) defined at workflow/envs/environment.yaml

The --use-envmodules flag allows to specify and load HPC modules (module load ...) in config/envmodules.yaml. This is especially useful on systems where AlphaFold or the alignment tools are available as environmental modules.

Querying the server

When the server is setup and running, it is possible to query it by sending an email to the server inbox specified in config/config.yaml. The email should be in plain text and the sequences should be pasted in the body as follows (no attachments):

TARGET field contains the name of the target
SEQUENCE field: contains one (monomer) or more (multimer) fasta sequences (see examples below)
REPLY field: the email address where results should be sent to. This does not have to be the same as the sender address
STOICHIOMETRY field: specify how many times each chain (starting from A) should be repeated in the model

Example: monomer

TARGET=targetID
SEQUENCE=GHMSKPQTQGLAKDAWEIPRESLRLEVKLGQGCFGEVWMGTWNGTTRVAIKTLKPGTMSPEAFLQEAQVMKKLRHEKLVQLYAVVSEEPIYIVTEYMSKGSLLDFLKGEMGKYLRLPQLVDMAAQIASGMAYVERMNYVHRDLRAANILVGENLVCKVADFGLARLIEDNEYTARQGAKFPIKWTAPEAALYGRFTIKSDVWSFGILLTELTTKGRVPYPGMVNREVLDQVERGYRMPCPPECPESLHDLMCQCWRKDPEERPTFEYLQAFLEDYFTSTEPQYQPGENL
REPLY=user@mail.com
STOICHIOMETRY=A1

Example: homodimer (A2 stoichiometry)

TARGET=targetID_homodimer
SEQUENCE=GHMSKPQTQGLAKDAWEIPRESLRLEVKLGQGCFGEVWMGTWNGTTRVAIKTLKPGTMSPEAFLQEAQVMKKLRHEKLVQLYAVVSEEPIYIVTEYMSKGSLLDFLKGEMGKYLRLPQLVDMAAQIASGMAYVERMNYVHRDLRAANILVGENLVCKVADFGLARLIEDNEYTARQGAKFPIKWTAPEAALYGRFTIKSDVWSFGILLTELTTKGRVPYPGMVNREVLDQVERGYRMPCPPECPESLHDLMCQCWRKDPEERPTFEYLQAFLEDYFTSTEPQYQPGENL
REPLY=user@mail.com
STOICHIOMETRY=A2

Example: heteromer

TARGET=targetID_heterotetramer
SEQUENCE=>subunit1|
MGSKKLKRVGLSQELCDRLSRHQILTCQDFLCLSPLELMKVTGLSYRGVHELLCMVSRACAPKMQTAYGIKAQRSADFSPAFLSTTLSALDEALHGGVACGSLTEITGPPGCGKTQFCIMMSILATLPTNMGGLEGAVVYIDTESAFSAERLVEIAESRFPRYFNTEEKLLLTSSKVHLYRELTCDEVLQRIESLEEEIISKGIKLVILDSVASVVRKEFDAQLQGNLKERNKFLAREASSLKYLAEEFSIPVILTNQITTHLSGALASQADLVSPADDLSLSEGTSGSSCVIAALGNTWSHSVNTRLILQYLDSERRQILIAKSPLAPFTSFVYTIKEEGLVLQAYGNS
>subunit2|
MRGKTFRFEMQRDLVSFPLSPAVRVKLVSAGFQTAEELLEVKPSELSKEVGISKAEALETLQIIRRECLTNKPRYAGTSESHKKCTALELLEQEHTQGFIITFCSALDDILGGGVPLMKTTEICGAPGVGKTQLCMQLAVDVQIPECFGGVAGEAVFIDTEGSFMVDRVVDLATACIQHLQLIAEKHKGEEHRKALEDFTLDNILSHIYYFRCRDYTELLAQVYLLPDFLSEHSKVRLVIVDGIAFPFRHDLDDLSLRTRLLNGLAQQMISLANNHRLAVILTNQMTTKIDRNQALLVPALGESWGHAATIRLIFHWDRKQRLATLYKSPSQKECTVLFQIKPQGFRDTVVTSACSLQTEGSLSTRKRSRDPEEEL
>subunit3|
MGVLRVGLCPGLTEEMIQLLRSHRIKTVVDLVSADLEEVAQKCGLSYKALVALRRVLLAQFSAFPVNGADLYEELKTSTAILSTGIGSLDKLLDAGLYTGEVTEIVGGPGSGKTQVCLCMAANVAHGLQQNVLYVDSNGGLTASRLLQLLQAKTQDEEEQAEALRRIQVVHAFDIFQMLDVLQELRGTVAQQVTGSSGTVKVVVVDSVTAVVSPLLGGQQREGLALMMQLARELKTLARDLGMAVVVTNHITRDRDSGRLKPALGRSWSFVPSTRILLDTIEGAGASGGRRMACLAKSSRQPTGFQEMVDIGTWGTSEQSATLQGDQT
>subunit4|
GCSAFHRAESGTELLARLEGRSSLKEIEPNLFADEDSPVHGDILEFHGPEGTGKTEMLYHLTARCILPKSEGGLEVEVLFIDTDYHFDMLRLVTILEHRLSQSSEEIIKYCLGRFFLVYCSSSTHLLLTLYSLESMFCSHPSLCLLILDSLSAFYWIDRVNGGESVNLQESTLRKCSQCLEKLVNDYRLVLFATTQTIMQKASSSSEEPSHASRRLCDVDIDYRPYLCKAWQQLVKHRMFFSKQDDSQSSNQFSLVSRCLKSNSLKKHFFIIGESGVEFC

REPLY=user@mail.com
STOICHIOMETRY=A1B1C1D1

Server responses

The server will respond as soon as it detects a new query by sending a confirmation to the sender address (not necessarily the REPLY address).

When the target has been modelled, it sends the models as PDB coordinates pasted in plain text to a number of emails (one per model) to the email address specified in the REPLY address (not the sender address).

Uploading results to a separate server

It is also possible to upload the AlphaFold outputs to a separate server for backup or so that users can access MSAs, pickle files etc. The access must be configured through public/private key and be passwordless (e.g. the server must be able to run rsync to the server without password prompt).

snakemake -j 1 --rerun-incomplete --cores 1 upload_msas # upload MSAs only
snakemake -j 1 --rerun-incomplete --cores 1 upload_models # upload models, pickle files