Improve SelectVariants rsID matching

[vdauwera]

Feature request

Tool(s) or class(es) involved

SelectVariants (all versions -- tested most recently with 4.0.9.0)

Description

SelectVariants has an option for excluding specific sites by rsID, --exclude-ids. Currently the matching is done by exact string matching. The limitation is that if I have a site with two rsIDs (as appears in the 1000 Genomes dataset), I can't filter against just one of them. For example, if I want to exclude this site from 1000G participant HG00096 (because it makes my synthetic data generator choke, if you must know):

3523879:16	68401338	rs554836707;rs56090907	CGTGCGCTGCT	CTGCT,C	100	PASS <snip>

I have to specify --exclude-ids 'rs554836707;rs56090907' in my SelectVariants command. I would like to be able to specify just --exclude-ids rs554836707 or --exclude-ids rs56090907 with the same result (the site is excluded). Right now if I do use just one like that, the site is not excluded because the pattern fails to match.

The reason I want to be able to do that is because there are several other similar sites that I need to exclude, so I make an array of rsIDs to exclude, e.g. provide

Array[String] = [
    "rs575450111",
    "rs151266607",
    "rs557010312",
    "rs564559134",
    "rs560466806",
    "rs140802196",
    "rs554836707",
    "rs56090907"
]

as an input to my WDL, which handles it in the command block as

--exclude-ids ${sep=' --exclude-ids ' exclusionList}

Right now I have to use

[
"'rs575450111;rs151266607'",
"'rs557010312;rs564559134'",
"'rs560466806;rs140802196'",
"'rs554836707;rs56090907'"
]

which is stylistically ugly and would not cover cases where only one rsID is annotated for whatever reason.

[davidbenjamin]

@MartonKN It looks like the relevant code is lines 713-5 of SelectVariants:

if (rsIDsToRemove != null && !rsIDsToRemove.isEmpty()) {
            compositeFilter = compositeFilter.and(new CountingVariantFilter(new VariantIDsVariantFilter(rsIDsToRemove).negate()));
}

and in VariantIDsVariantFilter

public boolean test(final VariantContext vc) {
        return includeIDs.contains(vc.getID());
    }

This seems like it should do the right thing, except that vc.getID() is a String, not a List<String>. @vdauwera can we assume that a VariantContext ID will not contain a semicolon, and therefore we can parse by splitting on ';'?

[davidbenjamin]

Closed by #5464.