incomplete citation information

[arw36]

Is the only citation information that which is provided in the provenance.csv? I was surprised to see many entries missing author information, as well as lots of bibliographic short hand that make it difficult/impossible to find the original source. Why aren't full references details provided?

I did try to download temporal.csv as well, but it wouldn't unpack on my mac.

[collinschwantes]

Hey Anna - thanks for pointing this out! I will look into the missing author information and pubmed ids and try to resolve them ASAP.

What would full reference details look like for you? Would that be all records are complete with the fields we currently have in the dataset or are there additional fields you think should be added?

[arw36]

Is there metadata for what each of the columns mean? In particular not sure what y'all require for "referenceText".

My interpretation for "referenceText" of publications is a minimum of Authors, Year, Title, Journal, Pages. Of course, links to the item or the item's more detailed citation information like PMID, dois, Jstor, Zenodo are nice, too.

[collinschwantes]

In the datapackage.json file there are descriptions of the fields in the provenance table. The description for reference text could be better.

"name": "provenance.csv",
            "path": "provenance.csv.gz",
            "profile": "tabular-data-resource",
            "format": "csv",
            "mediatype": "text/csv",
            "encoding": "UTF-8",
            "schema": {
                "fields": [
                    {
                        "name": "AssocID",
                        "type": "number",
                        "description": "Row number from current VIRION version"
                    },
                    {
                        "name": "HostOriginal",
                        "type": "string",
                        "description": "Host name from original dataset"
                    },
                    {
                        "name": "VirusOriginal",
                        "type": "string",
                        "description": "Virus name from original dataset"
                    },
                    {
                        "name": "Database",
                        "type": "string",
                        "description": "Source for the record. One of EID2, Shaw, HP3, GMPD2, PREDICT, OR GenBank"
                    },
                    {
                        "name": "DatabaseVersion",
                        "type": "string",
                        "description": "For static data, a citation. For dynamic data (e.g. Genbank) the access URL and a time stamp"
                    },
                    {
                        "name": "ReferenceText",
                        "type": "string",
                        "description": "A text description of literature sources"
                    },
                    {
                        "name": "PMID",
                        "type": "number",
                        "description": "PubMed identifiers for literature sources"
                    }
                ]
            }
        },

[arw36]

Ah, thanks for pointing to that. I agree the provided metadata doesn't give any further detail, and doesn't seem to be consistently adhered to. My inclination would be to refer to an industry-standard like Chicago style. Some example journal standards below:

• https://www.nature.com/sdata/submission-guidelines#references
• https://www.nature.com/sdata/submission-guidelines#data_citations
• https://esajournals.onlinelibrary.wiley.com/hub/journal/19399170/author-guidelines#references

another thing I find helpful is when empty cell are explicit that they don't have certain information. Such as PMID: "No PMID available"

[collinschwantes]

I'm going to look a little more closely at how the provenance data are compiled and figure out how we can improve the data. It would probably be helpful to look for a set of persistent identifiers that provide good coverage for all references (DOIs, Genbank accession numbers, PMIDS?).

@arw36 can you describe how you plan on using the references (even if its broad brush) so I can make sure the updated provenance data meets your needs?

[arw36]

Generally to find the original sources of host-parasite interactions. Since this isn't really viable through the text references, it requires going to the original database. With each database being different, I think this duplicates a lot of the point of VIRION already in trying to harmonize this into one system.

Secondly, I would use to remove some duplicates to quanitify research effort or # of publications per host-virus. e.g. the Adelaide river ephemerovirus in Bos taurus is listed as from "Gard GP, 1984, Aust Vet J, 61" via Shaw and "Gard et al. 1984" from Olival. By pure reference information alone, these aren't de-duplicated by code with out customisation.

For a subset analysis, to get the original references into my bibliographic software, with minimal citation information I could batch outsource this, but as it stands its a lot of manual looking or returning to the original database.