Assay assignments at sample level

[Steven Foltz]

Hi -- separate inquiry! Thank you again for this valuable resource and the recent updates with species and assay. Those are very helpful and enabled a great expansion of the database!

One minor issue I'm running into has to do with assay assignments at the sample level. A single sample may have multiple experiments derived from it, and those experiments may have had different assays. However, only a single assay has been assigned at the sample level in the database. This is coming up in the context filtering sample lists based on assay.

For example, there are multiple experiments derived from this sample, some with RNA-seq and some with ChIP-seq assays. https://www.ncbi.nlm.nih.gov/sra/?term=SRS2436922

From MetaSRA v1.8, there is only evidence of a single assay.

"SRS2436922": {

        "real-value properties": [],

        "assay": "ChIP-seq",

        "sample-type confidence": 0.9612915935073109,

        "mapped ontology terms": [

            "EFO:0000322",

            "CL:0000010"

        ],

        "sample type": "cell line",

        "species": "human"

    },

Is it possible to link assay to individual experiments? Or indicate that for a given sample there are experiments using different assays?

I'm also happy to file this as an issue on https://github.com/deweylab/metasra-pipeline if that would suit your workflow better.

Thank you!

Steven Foltz

[bernie...@gmail.com]

Hi Steven,

Thanks again for your interest and for pointing out this issue.  Ah, yes this is indeed a problem with our current schema.  Originally, we designed the MetaSRA to be only used for RNA-seq and we used the sample accession, rather than the experiment accession, as the sample identifiers in the MetaSRA. However, when we went to extend MetaSRA to other assays, this choice resulted in the problem you describe: certain samples may be assayed with multiple protocols.  Unfortunately, there's no quick-fix here; however, for a future release of the MetaSRA we may need to transition to experiment accessions, or at least update the "assay" field to be a list of assays rather than a single assay.  

I will look into doing this for the next release.  

Best,

Matt