Benchmarking single-cell RNA-sequencing protocols for cell atlas projects

ipediez commented 2 years ago

Project short name:

BenchmarkingSingleCellProtocols

Primary Wrangler:

Ami

Secondary Wrangler:

Ida

Associated files

Published study links

Paper: https://doi.org/10.1038/s41587-020-0469-4
Accessioned data: GSE133549

Key Events

[x] Convert published metadata to HCA spreadsheet
[x] Manually curate dataset to meet HCA metadata standard
[x] Collect any matrix and cell-type annotation files
[x] Upload sheet to validate metadata
[ ] Check linking using ingest graph validator
[x] Transfer raw files to ingest to validate data files
[x] Ask the Secondary Wrangler for an end-to-end review of the project. Ask the Expertise Wrangler to review specific tabs if needed
[ ] Submit dataset to Production
[ ] Complete the Export SOP
[ ] Convert project data to SCEA format following the SCEA conversion SOP if appropriate

ami-day commented 2 years ago

Requires a new library preparation method ontology term. Request is here: https://github.com/HumanCellAtlas/ontology/issues/107

Requires library preparation method info. from the authors (various methods). I have emailed the authors to ask for this.

ofanobilbao commented 2 years ago

Moving this back to Wrangling while Ami is away, as I am not sure if this is ready for secondary review or not, as I have not seen any posts asking for review and not all the boxes in the acceptance criteria are ticketed.

idazucchi commented 2 years ago

@gabsie and @ami-day to dicuss if this will be included in the DCP

Wkt8 commented 2 years ago

@gabsie what is the status of this?

ofanobilbao commented 2 years ago

@gabsie to investigate

gabsie commented 2 years ago

OK, on this project - let's just make a decision, this has been sitting here forever. The pipelines are no longer an issue. Let's publish the project with both raw and matrices. Any last issues from the rest of the wranglers? @ESapenaVentura @idazucchi @Wkt8

gabsie commented 2 years ago

OK - let's do it - both raw and matrices go into and we publish it. @ami-day - please proceed

idazucchi commented 2 years ago

Postponed to R21 @ami-day

ami-day commented 2 years ago

Requested data via NCBI Cloud Delivery

ami-day commented 1 year ago

064f96e3-2af9-43b0-967e-b86cdd876e79 63a86149-cfa5-417f-8e2f-90229a5a1fca 1af5c8e3-9c8f-4722-afee-f084f6c59f86 6c0f012d-7592-4345-827e-f9ed0d706013

gabsie commented 1 year ago

The majority of files has been downloaded. Ida will finish secondary review. This will probably not make it to release 21, but good to finalise it. @ami-day

ami-day commented 1 year ago

Synced data files. Graph validating

ami-day commented 1 year ago

Waiting on validation of metadata schema test data (cell suspension -> cell suspension as experiment design)

idazucchi commented 1 year ago

Hi ami, I'm done with the review. This is a large dataset with a lot of libraries and cell suspensions. I think we can simplify the way we model the experimental design and still be accurate by :

using the same cell suspension reference_sample as input for all the libraries and files to highlight that they are all using the same cell mix.
delting the libarary specific enrichment and adding a few lines to the library protocol description if needed. There is no enrichment that requires them

Project

GSE141469 is in the project but all the files and library are missing from the spreadsheet
some organs are missing
the organisms are missing
the ontology label and id for data curator are in the wrong fields
the publication url points to the abstract

Donor

the human PBMCs donor are 2 female and 2 male
human adult stage can't be used for non human donors
MDCK_donor should be female, species canis lupus familiaris

Collection

a protocol is needed to connect the cell line donors to the specimens

Specimens

the cell line specimens have no collection protocol

Cell line

you can add the publication information
there is no protocol to connect the cell lines to their specimens

Enrichment

reference_sample_processing_protocol could be fluorescence-activated cell sorting for DAPI- cells. You can use it connect together all the cells pooled for reference_sample and add "During FACS isolation, DAPI-positive cells were excluded to remove dead and damaged cells. " to the description

Cell suspension

I would add a cell suspension for the cell lines and use dissociation_protocol_cell_line for the dissociation protocol
reference_sample can have all the colon, PBMCs and cell line suspensions as input and reference_sample_processing_protocol for enrichment protocol

Library prep

library_preparation_protocol_C1 according to the supplemetary material this technique is barcoded. The GEO accession suggests R1 = BC (1-6), UMI (7-11) but I have not found anything to confirm that. The end bias should be 3' and the strand might be first.
all these techniques are barcoded according to the supplemetary material:
- library_preparation_protocol_ICELL8 these are the barcodes length according to geo R1 = BC (1-11), UMI (12-22); R2 = cDNA
- library_preparation_protocol_MARSseq these are the barcodes length according to geo R1 = cDNA; R2 = BC (1-6), UMI (7-14)
- library_preparation_protocol_QuartzSeq2 these are the barcodes length according to geo R1 = BC (1-15), UMI (16-23), R2 = cDNA
- library_preparation_protocol_gmSCRBseq these are the barcodes length according to geo R1 = BC (1-6), UMI (7-16); R2 = cDNA
- library_preparation_protocol_ddSEQ these are the barcodes length according to geo R1 = BC (8-14,25-31,42-48), UMI (1-7); R2 = cDNA

Sequencing protocol

sequencing_protocol_C1 should also be tag-based
sequencing_protocol_CELSeq2 should have HiSeq 3000 according to GEO
sequencing_protocol_ICELL8 should have HiSeq 4000 according to GEO
sequencing_protocol_gmSCRBseq should have HiSeq 2000 according to GEO
sequencing_protocol_inDrop should have HiSeq 4000 according to GEO

Sequence files

smartSeq2 is multiplexed so we need I1 and I2 to demultiplex, are they available?

Analysis protocol

the protocol description contains the numbered references from the publication
the computational method field is a copy of derivation of process and it contains double pipes
analysis_protocol_processedsays the data was normalised but the counts in the file don't look like they have been normalised. Is this the protocol applied to the data available from geo? If not it can be deleted

Analysis file

the content description for metadata files should not be gene expression matrix
you could add the matrix cell count
the analysis matrix for the metadata files should be analysis_protocol_raw

ami-day commented 1 year ago

Updated dataset based on review comments. Need to convert SRR9621775 to fastq. Waiting on test data test.

ami-day commented 1 year ago

Waiting on this PR: https://github.com/HumanCellAtlas/schema-test-data/pull/28

Wkt8 commented 1 year ago

Unblocked by cell-to-cell-suspension test data.

idazucchi commented 1 year ago

Lower priority than kidney datasets - will tackle this in release 26

gabsie commented 1 year ago

needs a look from another wrangler on the modelling - maybe @anu-shiva and @Wkt8

idazucchi commented 1 year ago

two additional issues:

[ ] smartSeq2 has both I1 and I2 and that is not allowed by the graph validation
[x] cell size selection for 10x only, apply to sequence file??

idazucchi commented 1 year ago

edit smartseq2_has_2-3_files

COUNT(f.`read_index`) as num_files, COUNT(DISTINCT f.`read_index`) as read_set
AND NOT num_files = read_set
RETURN p, "Smart-Seq2 protocols should contain between 2 and 3 files", labels(p)

? f is all files? I need just the ones linked to the process

gabsie commented 1 year ago

blocked by the graph validation issue

gabsie commented 1 year ago

Import form is filled

ebi-ait / hca-ebi-wrangler-central