GeneNetwork3 REST API for data science and machine learning

GeneNetwork3 is a light-weight back-end that serves different front-ends, including the GeneNetwork2 web UI. Transports happen in multiple ways:

1. A REST API
2. Direct python library calls (using PYTHONPATH)

The main advantage is that the code is not cluttered by UX output and starting the webserver and running tests is easier than using GeneNetwork2. It allows for using Jupyter Notebooks and Pluto Notebooks as front-ends as well as using the API from R etc.

A continuously deployed instance of genenetwork3 is available at https://cd.genenetwork.org/. This instance is redeployed on every commit provided that the continuous integration tests pass.

The system comes with some default configurations found in "gn3/settings.py" relative to the repository root.

To overwrite these settings without changing the file, you can provide a path in the GN3_CONF environment variable, to a file containing those variables whose values you want to change.

The GN3_CONF variable allows you to have your own environment-specific configurations rather than being forced to conform to the defaults.

Install GNU Guix - this can be done on every running Linux system.

There are at least three ways to start GeneNetwork3 with GNU Guix:

1. Create an environment with guix shell
2. Create a container with guix shell -C
3. Use a profile and shell settings with source ~/opt/genenetwork3/etc/profile

At this point we use all three for different purposes.

Simply load up the environment (for development purposes):

guix shell -Df guix.scm

Also, make sure you have the guix-bioinformatics channel set up correctly and this should work

guix shell --expose=$HOME/genotype_files/ -Df guix.scm
python3
  import redis

Check if guix and guix-bioinformatics channel are up-to-date with

guix describe

Containers provide full isolation from the underlying distribution. Very useful for figuring out any dependency issues:

guix shell -C --network --expose=$HOME/genotype_files/ -Df guix.scm

A guix profile is different from a Guix shell - it has less isolation from the underlying distribution.

Create a new profile with

guix package -i genenetwork3 -p ~/opt/genenetwork3

and load the profile settings with

source ~/opt/genenetwork3/etc/profile
start server...

Note that GN2 profiles include the GN3 profile (!). To roll genenetwork3 back you can use either in the same fashion (probably best to start a new shell first)

bash
source ~/opt/genenetwork2-older-version/etc/profile
set|grep store
run tests, server etc...

If you get a Guix error, such as ice-9/boot-9.scm:1669:16: In procedure raise-exception: error: python-sqlalchemy-stubs: unbound variable it typically means an update to guix latest is required (i.e., guix pull):

guix pull
source ~/.config/guix/current/etc/profile

and try again. Also make sure your ~/guix-bioinformatics is up to date.

See also instructions in .guix.scm.

These configurations should be set in an external config file, pointed to with the environment variable GN3_CONF.

• SPARQL_ENDPOINT (ex: "https://localhost:9082/sparql")
• RQTL_WRAPPER (ex: "~/genenetwork3/scripts/rqtl_wrapper.R")
• XAPIAN_DB_PATH (ex: "/export/data/genenetwork/xapian")

TMPDIR also needs to be set correctly for the R script(s) (previously there was an issue with it being set to /tmp instead of ~/genenetwork3/tmp)

This project has a number of utility scripts that are needed in specific cirscumstances, and whose purpose is to support the operation of this application in one way or another. Have a look at the [Scripts.md file](./docs/Scripts.md] to see the details for each of the scripts that are available.

NOTE: Do not create migration scripts manually. Use the processes indicated below.

The migration scripts for the authentication and authorisation system are in the migrations/auth folder in the root of the repository.

To create an new migration script for the, do:

$ yoyo new -m "<description of the migration>" ./migrations/auth/

The command will ask whether you want to save the migration configuration, e.g.

$ yoyo new --config=yoyo.auth.ini -m "testing a new migration"
Error: could not open editor!
Created file ./migrations/auth/20221103_02_HBzwk-testing-a-new-migration.py
Save migration configuration to yoyo.ini?
This is saved in plain text and contains your database password.

Answering 'y' means you do not have to specify the migration source or database connection for future runs [yn]:

If you specify y then a file named yoyo.ini will be created in your current working directory, and you can refer to it to avoid providing the ./migrations/auth explicitly.

Now you can open and edit the scripts to provide the appropriate SQL statements to update or rollback your schema.

To apply the migrations, you can do something like:

$ yoyo apply --database="sqlite:https:////tmp/test-auth.db" ./migrations/auth/

[20221103_01_js9ub-initialise-the-auth-entic-oris-ation-database]
Shall I apply this migration? [Ynvdaqjk?]: Y

[20221103_02_sGrIs-create-user-credentials-table]
Shall I apply this migration? [Ynvdaqjk?]: Y

[20221108_01_CoxYh-create-the-groups-table]
Shall I apply this migration? [Ynvdaqjk?]: Y

[20221108_02_wxTr9-create-privileges-table]
Shall I apply this migration? [Ynvdaqjk?]: Y

...

If you have previously initialised the yoyo config file, you can put the database uri in the configuration file and just provide it to avoid the prompt to save the configuration.

As a convenience, and to enable the CI/CD to apply the migrations automatically, I have provided a flask cli command that can be run with:

$ export FLASK_APP=main.py
$ flask apply-migrations

This expects that the following two configuration variables are set in the application:

• AUTH_MIGRATIONS: path to the migration scripts
• AUTH_DB: path to the sqlite database file (will be created if absent)

To enable you to run the OAuth2 server without HTTPS, you need to setup the following environment variable(s):

• export AUTHLIB_INSECURE_TRANSPORT=true: Allows you to run the Authlib server without HTTPS on your development machine.

In this section, we present some example request to the API using cURL to acquire the token(s) and access resources.

curl -X POST https://localhost:8080/api/oauth2/token \
    -F "[email protected]" -F "password=testpasswd" \
    -F "grant_type=password" \
    -F "client_id=0bbfca82-d73f-4bd4-a140-5ae7abb4a64d" \
    -F "client_secret=yadabadaboo" \
    -F "scope=profile group role resource register-client user introspect migrate-data"

Once you have acquired a token as above, we can now access a resource with, for example:

curl -X GET -H "Authorization: Bearer L3Q5mvehQeSUNQQbFLfrcUEdEyoknyblXWxlpKkvdl" \
    "https://localhost:8080/api/oauth2/group/members/8f8d7640-5d51-4445-ad68-7ab217439804"

to get all the members of a group with the ID 8f8d7640-5d51-4445-ad68-7ab217439804

or:

curl -X POST "https://localhost:8080/api/oauth2/user/register" \
    -F "[email protected]" -F "password=apasswd" \
    -F "confirm_password=apasswd"

where L3Q5mvehQeSUNQQbFLfrcUEdEyoknyblXWxlpKkvdl is the token you got in the Request Token section above.

(assuming you are in a guix container; otherwise use venv!)

To run tests:

export AUTHLIB_INSECURE_TRANSPORT=true
export OAUTH2_ACCESS_TOKEN_GENERATOR="tests.unit.auth.test_token.gen_token"
pytest

To specify unit-tests:

export AUTHLIB_INSECURE_TRANSPORT=true
export OAUTH2_ACCESS_TOKEN_GENERATOR="tests.unit.auth.test_token.gen_token"
pytest -k unit_test

Running pylint:

pylint *py tests gn3 scripts sheepdog

Running mypy(type-checker):

mypy --show-error-codes .

To spin up the server on its own (for development):

export FLASK_DEBUG=1
export FLASK_APP="main.py"
flask run --port=8080

And test with

curl localhost:8080/api/version
"1.0"

To run with gunicorn

gunicorn --bind 0.0.0.0:8080 wsgi:app

consider the following options for development --bind 0.0.0.0:$SERVER_PORT --workers=1 --timeout 180 --reload wsgi.

And for the scalable production version run

gunicorn --bind 0.0.0.0:8080 --workers 8 --keep-alive 6000 --max-requests 10 --max-requests-jitter 5 --timeout 1200 wsgi:app

(see also the .guix_deploy script)

IMPORTANT NOTE: we do not recommend using pip tools, use Guix instead

1. Prepare your system. You need to make you have python > 3.8, and the ability to install modules.
2. Create and enter your virtualenv:

virtualenv --python python3 venv
. venv/bin/activate

3. Install the required packages

# The --ignore-installed flag forces packages to
# get installed in the venv even if they existed
# in the global env
pip install -r requirements.txt --ignore-installed

Make sure that the dependencies in the requirements.txt file match those in guix. To freeze dependencies:

# Consistent way to ensure you don't capture globally
# installed packages
pip freeze --path venv/lib/python3.8/site-packages > requirements.txt

During development, there is periodically need to log what the application is doing to help resolve issues.

The logging system was initialised to help with this.

Now, you can simply use the current_app.logger.* logging methods to log out any information you desire: e.g.

from flask import current_app

...

def some_function(arg1, arg2, **args, **kwargs):
    ...
    current_app.logger.debug(f"THE KWARGS: {kwargs}")
    ...

You can get the genotype files from https://ipfs.genenetwork.org/ipfs/QmXQy3DAUWJuYxubLHLkPMNCEVq1oV7844xWG2d1GSPFPL and save them on your host machine at, say $HOME/genotype_files with something like:

$ mkdir -p $HOME/genotype_files
$ cd $HOME/genotype_files
$ yes | 7z x genotype_files.tar.7z
$ tar xf genotype_files.tar

The genotype_files.tar.7z file seems to only contain the BXD.geno genotype file.

To run QTL computations, this system makes use of the rust-qtlreaper utility.

To do this, the system needs to export the trait data into a tab-separated file, that can then be passed to the utility using the --traits option. For more information about the available options, please see the rust-qtlreaper repository.

The traits file begins with a header row/line with the column headers. The first column in the file has the header "Trait". Every other column has a header for one of the strains in consideration.

Under the "Trait" column, the traits are numbered from T1 to T where is the count of the total number of traits in consideration.

As an example, you could end up with a trait file like the following:

Trait	BXD27	BXD32	DBA/2J	BXD21	...
T1	10.5735	9.27408	9.48255	9.18253	...
T2	6.4471	6.7191	5.98015	6.68051	...
...

It is very important that the column header names for the strains correspond to the genotype file used.

The partial correlations feature depends on the following external systems to run correctly:

• Redis: Acts as a communications broker between the webserver and external processes
• sheepdog/worker.py: Actually runs the external processes that do the computations

These two systems should be running in the background for the partial correlations feature to work correctly.