Review curation SoP

[dosumis]

Ugur needs only:

1. Something to identify dataset - currently h5ad link
2. Author cell type fields

In future:
author cell type field present : T/F (update SOP - there should be a row with blank author cell type field(s)
To deal with version changes, need CxG Link

Editors need

• At least one human readable title for dataset.

Details https://github.com/Cellular-Semantics/CL_KG/blob/main/docs/dataset_curation_guidelines.md

1. DataSet identification:

We have 7 fields:

Dataset (individual datasets within larger group):
Description: The specific name of the dataset being curated within a larger dataset group.
Example: "Single cell transcriptional and chromatin accessibility profiling redefine cellular heterogeneity in the adult human kidney - ATACseq"

Full name dataset (top of page):
Description: The full descriptive name of the dataset that should be used for documentation and display.
Example: "Single cell transcriptional and chromatin accessibility profiling redefine cellular heterogeneity in the adult human kidney"

CxG Link:
Description: The CellxGene link to access the dataset.
Example: "https://cellxgene.cziscience.com/e/13a027de-ea3e-432b-9a5e-6bc7048498fc.cxg/"

h5ad link:
Description: The direct link to the .h5ad data file of the dataset.
Example: "https://datasets.cellxgene.cziscience.com/dabd979f-cc50-4526-81f3-8bc6c673ca36.h5ad"

Reference_DOI:
Description: The DOI reference for the associated publication(s) for the dataset.
Example: "DOI: 10.1038/s41467-021-22368-w"

Study Short Name:
Description: The shortened name or acronym of the study associated with the dataset.
Example: "Muto et al. (2021) Nat Commun"

CxG Dataset Collection X:
Description: The CellxGene link to the collection where the dataset is stored.
Example: "https://cellxgene.cziscience.com/collections/9b02383a-9358-4f0f-9795-a891ec523bcc"

Do we need them all?

Use cases:

• Curators need something readable to work with to know what they've curated
• Ugur needs some specific key to look up. It may also be useful to have 2 keys, one human readable one not, to cross check. Right now he is using the h5ad link only. h5ad link is sensitive to version and might change.
• A record of what's been curated (although this can also be generated from reports)
• Loading dataset where there is no author cell type category but there is CL annotation

Not needed:

• DOI not needed as can get it from CxG link
• study short name
  CxG Dataset Collection <-- keep because helpful in debugging

h5ad should be the latest version which may not be the same as the CxG dataset link.

content

Suggestion:

• Remove 'content' column and include only 'author cell type field' column. Having another column with only the entry 'cell type' would be pointless.

[ubyndr]

Hi @JABelfiore , could you please review this issue and let me know which fields are required from a development perspective? I plan to remove any unnecessary fields.

For development, I only require the following columns:

• Content
• CxG Link
• Author Category Cell Type Field Name

[JABelfiore]

Hi @ubyndr I need the following: Dataset (individual dataset name), Full name dataset, author cell type field name, CxG dataset collection, is the dataset normal? and stage. Thanks!