search

x-atlas-consortia / hra-pop

HRApop
MIT License
0 stars 1 forks source link
human-reference-atlas

readme

HRApop

This repository has the code, input data, and draft results for HRApop construction. Final versions of the HRApop atlas are published to https://lod.humanatlas.io/graph/hra-pop (and https://lod.humanatlas.io/graph/hra-pop-lq for the lower-quality version). An ER Diagram of the resulting graph is here.

Construction Algorithm

Requirements

To run the construction algorithm, you will need the following installed:

  1. A unix-like environment (Linux, WSL2 / Ubuntu For Windows, or Mac (untested))
  2. Node.js v18+
  3. jq (a lightweight command-line JSON processor)
  4. Java 11 (for blazegraph-runner)
  5. Docker (optional)

Setup

  1. Install node dependencies via npm ci, which also installs blazegraph-runner into node_modules/.bin for querying and reports.
  2. Install Java 11 and jq. On Ubuntu, this is typically done via sudo apt install openjdk-11-jdk jq

Input

Each HRApop version is defined in a subdirectory of the input-data directory by version. A config.sh file is used to configure the sources and settings for the HRApop construction workflow.

In addition to the config.sh file, the input-data/$VERSION directory will have a few things:

Running

To start a workflow run, check the constants.sh to ensure it's including the right config.sh for your version. Then run ./logged-run.sh which will run the whole workflow and place a log.txt file in the correct subdirectory of output-data.

Construction Workflow

Script Description
00-setup-environment.sh Additional environment setup (installs blazegraph-runner)
05-build-deprecated-cell-summaries.sh Get cell summaries via old CTPop method
05-build-deprecated-corridors.sh Get manually generated corridors (to eventually be replaced by a web service)
05-build-deprecated-registrations.sh Get registrations (hra-dataset-graphs) via old CTPop method
10-process-registrations.sh Combine registrations and compute collisions and euclidian distances
20-process-cell-summaries.sh Combine cell summaries and unflatten dataset metadata to hra-dataset-graph.jsonld format
30-process-publications.sh Unflatten publication metadata to hra-dataset-graph.jsonld format
40-normalize-dataset-graph.sh Combine and deduplicate registrations, cell summary, and publication datasets into a single full-dataset-graph.jsonld
45-split-dataset-graph.sh Split full-dataset-graph.jsonld into atlas (3 diamond datasets), atlas-lq (2 diamond, extraction site + cell summary datasets), test (1 diamond, extraction site OR cell summary datasets), and non-atlas datasets (anything not 3 diamond as a flat csv for tracking/improving)
50-generate-as-cell-summaries.sh Compute the AS Cell Summaries for Atlas and Atlas LQ
55-generate-extraction-site-cell-summaries.sh Compute Cell Summaries for Extraction Sites for Atlas, Atlas LQ, and Test data (using Atlas and Atlas LQ AS Cell Summaries)
58-compute-cell-summary-similarities.sh Compute cell summary similarities between all cell summaries generated.
60-enrich-dataset-graphs.sh For Atlas, Atlas LQ, and test data, add collisions, corridors, and cell summaries to there dataset-graphs to generate *-enriched-dataset-graph.jsonld files
70-create-internal-blazegraph.sh Load *-enriched-dataset-graph.jsonld files, extraction site distances, and cell summary similarities into a Blazegraph db for querying.
75-run-reports.sh Run reports against the generated blazegraph db using Atlas and Atlas LQ
75x-run-remote-reports.sh Run reports against Atlas and Atlas LQ using the HRA-KG SPARQL server at https://lod.humanatlas.io/sparql.
80-publish-results.sh Compile the data for publication, including Atlas and Atlas LQ enriched dataset graphs, non-atlas-dataset-graph.csv (for tracking/improving datasets), and the reports generated against the atlases.

Output

Data is compiled to output-data/$VERSION.

Atlas (3 Diamond Datasets):

Atlas LQ (2 Diamond Datasets):

All Non-3 Diamond Datasets: