Is it possible to ignore indels when reconstructing haplotypes using haplotypcaller? If not, would setting --PCR-indel-model to hostile be the best way to downplay indels?
That sounds reasonable. You could also increase the minPruning parameter to get rid of any graph paths that are very weakly supported. It will reduce your sensitivity in low-coverage regions, though.
Why do you want to ignore indels?
Along with your idea, you can also set a very low value for indelHeterozygosity.
Thanks very much for your comment.
Some sequencing technologies like iontorrent produce very high of false positive rates for indels. My thinking is, a random scattering of false indels across reads which contain a true heterozygous SNV will produce a high number of haplotypes. This may result in the loss of informative reads that do not fit the most probable haplotypes because of these false indels. Perhaps my thinking here is wrong?