Right now the BAM based interface of the consensus pileup generator encodes all reads in the input files. But it's useful to limit encoding to certain regions.
E.g. there could be regions to avoid (e.g. regions with known variations or lots of error)
E.g. specific regions to be used for training and validation and test
Regions can be passed using a BED file (the parser already exists in VW).
So I think the interface to the tool can have the following options, and each will take a BED file as input. We only need to support this for the case where BAM is supported as input (not where directories are supported as input).
--train-regions
--val-regions
--test-regions
--ignore-regions
Right now the BAM based interface of the consensus pileup generator encodes all reads in the input files. But it's useful to limit encoding to certain regions.
E.g. there could be regions to avoid (e.g. regions with known variations or lots of error)
E.g. specific regions to be used for training and validation and test
Regions can be passed using a BED file (the parser already exists in VW).
So I think the interface to the tool can have the following options, and each will take a BED file as input. We only need to support this for the case where BAM is supported as input (not where directories are supported as input).
--train-regions
--val-regions
--test-regions
--ignore-regions