HCA Ingest Service Graph Validation Suite

What is this useful for in the scope of the HCA:

1. Enables data wranglers to visually analyze the relationships inside a submission to look for inconsistencies.
2. Provides an automated graph validator for which to create tests using step 1 and can be run fully containerized.

Features

The suite is divided in two separate, extensible parts:

• hydrators enable users to import and populate data into a graph database. The reason not to call them importers is import is a reserved keyword in Python and from importers import importer is a bit confusing. 😵

• actions provide different tools to work with the generated graph. The first and most important is to run a series of tests to validate the constraints Data Wranglers want to impose on submissions. Another action is generating reports and extracting statistics from the graph to send to the submitters. Any other actions can be implemented to extend the suite.

Functionality

So far, the functionality planned is as follows (WIP items are still not fully implemented):

• Hydrators:

  • Ingest Service Spreadsheet.
  • Ingest Service API Submission.
  • BioSamples API (WIP).

• Actions:

  • Opening an interactive visualizer to query the graph.
  • Running tests on the graph.
  • Generating reports for the graph (WIP).

Installation

The Graph Validator Suite requires docker running in the host machine.

From the git repo

git clone git@github.com:ebi-ait/ingest-graph-validator.git
cd ingest-graph-validator
python -mvenv .venv
source .venv/bin/activate
pip install -e .

Usage

Step by step usage for running locally

1. Ensure Docker is installed and running

2. Create a file for the GCP credentials in your HOME folder: ~/.secrets/gcp_credentials and copy the secrets from AWS Secrets Manager under the following key: ingest/{ENV}/gcp-credentials.json. Where ENV could be one of the following values:

   1. dev for the dev environment
   2. staging for the staging environment
   3. prod for the production environment

3. run neo4j locally (in another terminal session or with -d (detached) flag)

docker run -p7687:7687 -p7474:7474 --env NEO4J_AUTH=neo4j/password --env=NEO4J_ACCEPT_LICENSE_AGREEMENT=yes neo4j:3.5.14-enterprise

4. export INGEST_GRAPH_VALIDATOR_INGEST_API_URL=https://api.ingest.archive.data.humancellatlas.org/

   • If you wish to run the graph validator against a different environment, you can specify the URL to that here (e.g. http://localhost:8080)

5. Initialize the database backend and enables a frontend visualizer to query the database, in http://localhost:7474 by default by executing this in the command line: ingest-graph-validator init

6. Import a spreadsheet:

   • ingest-graph-validator hydrate ingest <sub_uuid> (via ingest)
   • ingest-graph-validator hydrate xls <spreadsheet filename> (via a spreadsheet)

7. Go to http://localhost:7474 in a browser to open the frontend.

   • Username: neo4j
   • Password: password

8. You can then start writing cypher queries in the input field on top of the web frontend to visualize the graph. For example:

   MATCH p=(n) RETURN p

   Will show the entire graph. Keep in mind this will crash the browser on huge datasets.

9. Run tests

   • ingest-graph-validator action test <path_to_tests>
   • e.g ingest-graph-validator action test graph_test_set

Running as a queue listener

It is possible to run the graph validator so that it listens to a queue on RabbitMQ that receives submission UUIDs. Once a message is received from the queue the hydrate and action commands are ran for the given submission UUID. This is how the graph validator is ran in the Ingest k8s infrastructure

ingest-graph-validator action ingest-validator graph_test_set

The above command runs the listener for the graph_test_set

Running queue listener locally against locally running ingest

1. Ensure you have a locally running and populated Ingest Mongo DB
2. Make sure ingest-core and rabbitMQ are running
3. export INGEST_GRAPH_VALIDATOR_INGEST_API_URL=http://localhost:8080
4. docker run -p7687:7687 -p7474:7474 --env NEO4J_AUTH=neo4j/password --env=NEO4J_ACCEPT_LICENSE_AGREEMENT=yes neo4j:3.5.14-enterprise
5. ingest-graph-validator action ingest-validator graph_test_set
6. Trigger graph validation via:
   • Run the UI locally and trigger through the submission page
   • Or curl -X PUT http://localhost:8080/submissionEnvelopes/<submission_id>/graphValidationRequestedEvent

Graph Validator Architecture within Ingest

sequenceDiagram

    participant UI
    participant c as "ingest-core"
    participant gv as "ingest-graph-validator"
    participant q as RabbitMQ
    participant st as "ingest-state-tracking"

    gv->>q: Listen to queue
    UI->>c: PUT /submissionEnvelopes/{id}/graphValidationRequestedEvent
    c->>st: Request change of state to GRAPH_VALIDATION_REQUESTED
    activate st
    st-->>c: Commit change of state to GRAPH_VALIDATION_REQUESTED
    c->>q: Add graph validation message to queue
    q->>gv: Pick up message from queue
    activate gv
    gv->>c: PUT /submissionEnvelopes/{id}requestGraphValidating
    c->>st: Request change of state to GRAPH_VALIDATING
    activate st
    st-->>c: Commit change of state to GRAPH_VALIDATING
    note left of gv: Begin graph validation
    gv->>c: PATCH /{entity_type}/{id} update graphValidationErrors on each entity
    gv->>c: PUT /submissionEnvelopes/{id}/requestGraphValid or /submissionEnvelopes/{id}/requestGraphInvalid
    c->>st: Request change of state to GRAPH_VALID or GRAPH_INVALID
    activate st
    st-->>c: Commit change of state to GRAPH_VALID or GRAPH_INVALID

Loading

More help

The Graph Validator Suite uses a CLI similar to git. Running a command without specifying anything else will show help for that command. At each level, the commands have different arguments and options. Running any subcommand with -h or --help with give you more information about it.

The root level commands are:

• ingest-graph-validator init starts the database backend and enables a frontend visualizer to query the database, in http://localhost:7474 by default.

• ingest-graph-validator hydrate shows the list of available hydrators.

• ingest-graph-validator actions shows the list of available actions.

• ingest-graph-validator shutdown stops the backend.

Extra stuff

Useful cypher queries

Show all nodes and relationships

MATCH p = (n)
RETURN p

Show all nodes and relationship excluding some

MATCH p = (n)
WHERE NOT n:LABEL AND NOT n:LABEL
RETURN p

Expands paths from a node

This one will be shown with an example. The example selects the donor CBTM-376C from Meyer's Tissue Stability dataset, and expands the paths to show all biomaterials, processes and files linked to it.

Note: Make sure to strictly define only one node to use as the source, otherwise it will be confusing.

Note: You have to be careful not to include nodes that would link your path to another one. For example, protocol or project are linked to more than one experimental design.

The first two lines are used to select one single node from which to expand. The third line expands the path using these parameters:

1. n, the starting node or nodes (preferably one for your first queries).
2. "", the relationship filter. We are not filtering by any relations in this query.
3. "-project|-protocol", the label filter. We are excluding (hence the minus sign) any nodes with the labels project or (that is represented by the |) protocol.
4. 0 is the minimum depth. Normally 0. Otherwise the starting nodes get excluded.
5. -1 is used to determine the maximum depth for the path expansion. -1 means no limit. If you would set a 1 here, the result would be the CBTM-376C donor and its first level neighbours.

MATCH (n:donor_organism)
WHERE n.`biomaterial_core.biomaterial_id` = "CBTM-376C"
CALL apoc.path.expand(n, "", "-project|-protocol", 0, -1) YIELD path
RETURN path

List of Ingest submission IDs

• KidneySingleCellAtlas: d5410c6e-612d-421a-a66f-2de5e04dd050
• HumanColonicMesenchymeIBD: c51efa69-a348-46d4-9997-c7a11915d53b
• scRNAseqSystemicComparison: d1610c4a-76c6-4b69-af63-c74af869fa75
• TissueStability: fd52efcc-6924-4c8a-b68c-a299aea1d80f
• demo HPSI human cerebral organoids: 6f33a6c6-170d-460b-bbba-9668e9d77aaf
• Mouse Melanoma: 4756b97c-3666-4ff0-9c48-611dc06b7740
• Multiplexed scRNA-seq with barcoded antibodies: 6931d821-9cd2-4ae1-acf7-62cd3b90082e
• Reprogrammed_Dendritic_Cells: 8c45a848-ab32-4928-8dd6-567b75eaf7e1
• HPSI human cerebral organoids: fce97270-fce0-4744-8a4e-a93d95521852
• Tabula Muris: 14df1f92-155c-4da2-97fc-85601dee64da
• snRNA-seq_for_human_retina: 7d515499-b5b6-4083-95cd-6a393344865e
• Healthy and type 2 diabetes pancreas: c81f7d54-a27f-4212-a6df-88dde947f7cc
• Single cell transcriptome analysis of human pancreas: 5cd8b827-e4fa-4ffa-86f5-f0f8302baee9
• SingleCellLiverLandscape: 1e2601a5-8938-446c-bbb3-1f37a84b11da
• WongAdultRetina: 668791ed-deec-4470-b23a-9b80fd133e1c
• Single cell RNAseq characterization of cell types produced over : 0948a727-228f-4cfc-857e-6243c6aed08d
• Kidney biopsy scRNA-seq: 588281cb-93e2-4888-b75e-5cee5e39fc7f
• BM_PC: 67abd7a1-644b-48d0-b8b8-251d8195a742
• Human Hematopoietic Profiling: 7dbcf5ae-f8d7-487c-a3d4-794d8639a1e2
• Mouse Endoderm Project: 0fb44736-c50f-49ab-bcdc-0985e596b955
• Drop-seq, DroNc-seq, Fluidigm C1 Comparison: bf3116e5-1af1-46c2-8bbd-44dac49d1e7f
• 1M Immune Cells: 85e72912-9f91-4489-8169-3b43cc65a16a
• 1M Neurons: 064d36ca-ea4d-428f-b30f-0cf5e5350a9d
• MouseGastrulationAtlas: 19c5c89a-7bfb-4b4d-9ff5-3ef009304ba5
• Fetal/Maternal Interface: 46fe8bd7-329b-4f09-b227-5ee48c109c16
• Tissue stability: 02e89f20-84c8-4daa-aaeb-80f4a85733ff
• CD4+ cytotoxic T lymphocytes: 3bb965cb-2a6f-4802-a4f2-52d0423e6ecc
• Diabetic Nephropathy snRNA-seq: a69c731b-9db7-45e3-938f-daf2e4636a6e
• HDCA-Sweden-10x: 071fc37f-1a01-4cf7-a6f5-2662c42c12b6
• HumanTissueTcellActivation: 8b5feb5e-9039-4c54-9e79-053e490c141a
• HumanMousePancreas: 7b4cd093-bfa5-477e-9c95-69bafc1cb6bf

Credits

This package was created with Cookiecutter.