Added support for mpmap multi-mapped reads in the combined format (see --agglomerate-alns option in mpmap). Note that the graph needs to be an exon only splice graph in order for this to work properly.
Re-added read based clustering and optimized it. This was needed for the multi-mapping support.
Reduced memory usage during inference by using a more sparse probability representation.
Added use of haplotype frequencies to the inference model.
Improved mixing of Gibbs sampler by using multiple parallel chains.
Fixed probability normalization bug when estimating the haplotypes for transcripts.
Changed default mapq filter added in previous PR to 1. The mapq filtering actually decreased the accuracy a bit when looking at the results in more details.
Added probability precision threshold option. Used to collapse similar probability vectors and filter output.
Only haplotype combinations with a posterior above the precision threshold are now written to the output. This is in order to scale better for high ploidies.
This PR contains the following improvements:
--agglomerate-alnsoption in mpmap). Note that the graph needs to be an exon only splice graph in order for this to work properly.