Proccaserra patch 1

[proccaserra]

No description provided.

[floradanna]

I believe ENA adaptor expect the comment "STUDY_TYPE" = controlled vocabulary from ENA.

{
"name": "STUDY_TYPE",
"value": "Transcriptome Sequencing"
}

[proccaserra]

we can simplify this as the value can be inferred from the library information: ie. if RNA-SEQ, then STUDY_TYPE = "Transcriptome Sequencing" as default value.

This is something that I discussed many years ago with ENA , we can follow-up with Dipayan and April

[bedroesb]

We should discuss id some of these are wanted blockers or not:

1. The PTX file is not strict JSON: it contains NaN values, e.g. line 6906. Jackson/strict validators are likely to reject that before any adaptor logic.

2. The service model expects a wrapper: IsaJson has only investigation, while the PTX file is the investigation object at root (line 2). The CLI can handle this only with --investigation-is-root; direct Java adaptors will not.

3. Java value deserialization will break: ParameterValue.value is a Value object, not a scalar (ParameterValue.java, Value.java). PTX uses scalar values like "PAIRED" and "MGI" around line 7699.

4. The target repo comment key is wrong for MARS CLI: PTX uses "target repository" (line 6870); CLI expects exactly target_repository (target_repo.py, isa_json.py). So CLI validation/filtering will fail.

5. ENA metadata is incomplete. study-level ENA study creation requires STUDY_TYPE/existing_study_type; PTX assay comments only have target repo. Experiment creation then requires DESIGN_DESCRIPTION, LIBRARY_SELECTION, PLATFORM, and INSTRUMENT_MODEL; PTX has library layout/strategy/source, but not those exact normalized keys.

6. Run XML will fail on file metadata: PTX RNA-seq files use Checksum Type and Checksum (line 7306); WebinRunXmlCreator on study-level looks for exact comments file type and file checksum and hardcodes MD5.

7. Duplicated IDs/names are risky. PTX has 12 sources/samples but only 6 unique source/sample IDs, and the RNA-seq assay repeats acquisition and extract IDs. BioSamples maps by source/sample name/id and will overwrite accessions; ENA run/experiment creation dedupes by process/library id, so half the repeated RNA-seq runs can disappear.

[proccaserra]

+1 @bedroesb , fixing those issues now, will send a patch. thx for the detailed reporting.

• i. point 5 & 6 are handled correctly by ISA2SRA.py converter but it seems the MARS-cli does not support the information associated with the ISAconfiguration and expect INSDC style attribute tag.

As this is defined in the ISA configurations, we may need to add a dictionary to store alternative labels to avoid these types of failures.

• ii. Alignment needed:
  the SRA XSD requirements regarding checksums requires 2 attributes:
• 'checksum_method', the insdc key which actually accepts 2 possible values {"MD5", "SHA256"} . The restriction on "MD5" seems ENA specific. PrecisionTox (via Globus) relies on SHA hashing)
• 'file_checksum', the insdc key for providing the actual hash value.