Add support for reporting the hg19/37 position (if available) in ncbi_snp_query_api()

[jooolia]

As brought up in #92. It would be nice to add support for reporting the hg19/37 position (if available) in ncbi_snp_query_api().

[sinarueeger]

Test

  aa <- ncbi_snp_query("rs1610720")
  expect_equal(aa$Chromosome, "6")
  expect_equal(aa$BP, 29793285) ## on GRCh38
  expect_equal(aa$BP, 29761062) ## on GRCh37

[arjanboltjes]

I needed this function to work with hg19/37, so I had a look at your ncbi_snp_api.R code and the NCBI api output of one of the SNPs in your examples (rs420358), and it seems to me line 8 in your code decides whether you get assembly 38 or 37.
Changing line 8 to:
if (record$is_ptlp == 0 & length(record$placement_annot$seq_id_traits_by_assembly) > 0) {
gives me hg37 output.
I ran this for a list of SNPs of interest that I'm working on, of which I know the chr and bp position, and it nicely yields the hg19/37 output that I would expect, instead of the default hg38 output of the function in your package. The output variable also yields GRCh37.p13 as value.

Looking at some more NCBI api output, within the primary_snapshot_data that you pull the info from, there sometimes are more entries besides hg19/37 and hg38, but as far as I saw these don't pass the second argument of your if call in line 8.

If the above turns out to be correct - I did not thoroughly look into this any further than the above - perhaps you can add an argument to the function this way where you can toggle between hg19/37 and hg38?

As an aside, running this function, especially for a vector of variants, relatively often gives me an error:
Error in curl::curl_fetch_memory(url, handle = handle) : Error in the HTTP2 framing layer
Is this something you see more often with the default function?

[sinarueeger]

@arjanboltjes thanks for your input! Regarding Error in curl::curl_fetch_memory(url, handle = handle) : Error in the HTTP2 framing layer - can you replicate the error with a particular set of SNPs, or does it occur randomly? In any case, could you file an issue for that?

[VangelisTheodorakis]

Yes this would be super useful! Should we expect a fix soon?

[cruzpazv]

I made a fork and try to solve this problem, check in: https://github.com/cruzpazv/rsnps/blob/master/R/ncbi_snp_api.R
By default ncbi_snp_query as version GRCh38. For version GRCh37, add to the function like this: ncbi_snp_query(snps, version = "37").

[jooolia]

Thanks @cruzpazv we will try to have a look at your solution this week.

[sinarueeger]

Thanks @cruzpazv. Could you make a pull request so we can easily check the changes you made in your fork?