Exploring the Data Citation Corpus.
A trusted central aggregate of all data citations to further our understanding of data usage and advance meaningful data metrics.
The data is available by request, this repository documents my explorations of the data.
to_sql.php parses the JSON data and outputs SQL statements so we can construct a simple SQL database to explore the data.
The data comes from two sources, DataCite and the Chan Zuckerberg Initiative (CSI):
SELECT COUNT(id), sourceId FROM citation GROUP BY sourceId;
The data repositories that are cited are identified by local UUIDs, so it is non-trivial to figure out which repository is which. Here is a list so far:
| Repository | id | citation count |
|---|---|---|
| GenBank | 00363b65-f3ef-4fa9-8255-23ab269f4930 | 3755354 |
| PDB | 87646104-e5ef-494b-b2f3-a46c9572e003 | 1729783 |
| SNP | 6087b2e9-ecbf-4898-8047-5f484f1bce2f | 890431 |
Publishers of the journals in which citations were found are also identified using UUIDs, the top twenty of these are listed below, and names can be determined by comparing to the chart on the corpus dashboard.
| name | citation count | publisherId |
|---|---|---|
| 2136164 | e566bc45-b8bc-430c-ab2c-9c224e1c6f21 | |
| 1029617 | ec75ceb1-215c-4376-aa1c-4b39d15dc069 | |
| 938870 | 9ead11e4-bd7d-4c91-aff0-cb962676520a | |
| 768385 | bf7ba43c-7a3e-43e3-a9c2-6ed5b6fb6303 | |
| 704427 | 08d58a61-189f-4316-892b-908a1832603d | |
| 635059 | babceab8-4440-4c65-ad12-24784190dbae | |
| 315654 | 602471f4-3d02-45f7-9d59-661471761299 | |
| 312135 | af7d8efb-1a44-4a02-9d5b-29ceb6878117 | |
| 277952 | ||
| 276263 | 37fa820b-d158-43b4-8f67-e0c2f7364d35 | |
| 199813 | 55506166-9f8d-4685-967d-c71c7af956b7 | |
| 171526 | 21c1aa14-7ac4-4ccb-8fdc-8f7e3ab047a9 | |
| 147908 | 2189510e-6e8f-410c-bf2a-a92319d51b0e | |
| 114627 | faca9ac2-2c88-4277-acdd-0a1177c10094 | |
| 98882 | deba021e-5d63-48af-82b5-673c6507a03e | |
| 97239 | dba2ef73-893b-4c93-9123-ea3429d6c983 | |
| 92100 | cfd487dd-9342-49ec-b93a-a044da079368 | |
| 90016 | bd7beb5b-5e4d-4c9f-b99d-944bc8cd5bf3 | |
| Pensoft | 80907 | 9d72fbd4-0a14-4ee8-bac5-75ec06ababf7 |
| 80376 | c6e65534-0e8c-495f-99ed-04ee78761d3c | |
| 60503 | d2c56596-551e-4f1e-81e6-d7bafe1670f8 |
The Protein Data Bank (PDB) has 1,729,783 citations in the corpus. There are 177,220 distinct PDB identifiers cited.
SELECT DISTINCT UPPER(subjId)
FROM citation
WHERE repositoryId = '87646104-e5ef-494b-b2f3-a46c9572e003';
Running these through pub_clean.php 31,612 (18%) do not match the PDB pattern /^[0-9][A-Za-z0-9]{3}$/.
I downloaded a list of all PDB identifiers from https://files.wwpdb.org/pub/pdb/holdings/current_file_holdings.json.gz, and then loaded those identifiers into the table identifier.
SELECT COUNT(id) FROM identifier WHERE namespace='pdb';
There are 216,225 distinct PDB identifiers.
We can compare the PDB identifiers in the corpus with the actual PDB identifiers:
SELECT COUNT(citation.id) FROM citation
INNER JOIN identifier
ON UPPER(citation.subjId) = identifier.id
WHERE repositoryId = '87646104-e5ef-494b-b2f3-a46c9572e003' AND identifier.namespace = 'pdb';
This finds 1,233,993 PDB identifiers, which is 71% of the total in the corpus. Hence a little under a third of the PDB citations appear to be erroneous.
We can look at some mistaken identifiers:
SELECT citation.id, UPPER(citation.subjId), identifier.id
FROM citation
LEFT JOIN identifier ON UPPER(citation.subjId) = identifier.id
WHERE repositoryId = '87646104-e5ef-494b-b2f3-a46c9572e003'
AND citation.subjId LIKE "1%"
AND identifier.id IS NULL
LIMIT 100;