Variant calling on custom amplicons with extremely high coverage

[memento]

Hi,
I have a hard time obtaining variants from a bam file constructed using amplicon sequencing. When I load bam's on IGV I can see loads of reads (well, there is a limit of how much You can really see but the depth show correct coverage). I have 30.000-1.500.000 reads (identical ones) per each bam file.
I tried to use GATK with specific region bed file (thought that it will speed it up) but it takes forever (already 5h+! for 6 regions of ~250bp).
Tried samtools mpileup (multiple bam's at once) but I cannot find any of known variants (visible under IGV).
Shall I use it one-by-one on each bam (and combine?)?
Thanks!

[swbarnes2]

Sounds like downsampling would make your life easier.

Removing duplicates would be one quick way of accomplishing this. But there are other ways. Picard, for one, can downsample a .bam. You can also do a quick and dirty version yourself, by grepping for reads, say, from one tile.

[me_myself_andI]

I would second the downsampling approach.

The removal of duplicates can in theory and depending on your setup introduce some biases. For example if your looking at subpopulations in viral or bacterial sequencing (i.e. not a 'simple' diploid genome) you might end up with only a handful of reads after duplicate removal, and those will not represent the actual 'allele' frequencies.