MHC region exclusion

[xiangzhu]

We often exclude MHC region when estimating LD and performing further downstream analyses.

It would be nice to do this MHC exclusion for people.

The idea is to have the following simple data frame inside ldshrink, and a flag like remove_mhc=TRUE:

> HLA.hg19
    chrom start.base end.base
HLA     6   29719561 32883508
> is(HLA.hg19)
[1] "data.frame" "list"       "oldClass"   "vector"

Similarly we may want to remove centromeres as well.

Both MHC and centromeres data frames can be found here: https://github.com/smgogarten/GWASTools/tree/master/data.

This issue is also related to Issue #14, since they share the same idea: if we cannot estimate something reliably, then probably better no to estimate them at all.

Another very detailed note: pay attention to the sources of these regions -- e.g. if these regions are pulled out from UCSC, then they are 0-based.