miss out part of chimeric reads

[jiangst]

@mwyczalkowski Hi,
Thanks for providing the handy python script to detection of chimeric transcripts from scRNA-seq. It's a smart way to find the overlap CB:UB in 2 position. But it seems that the discover_chimeric_transcripts.py script miss out part of chimeric reads(one read mapped to 2 chromosomes/genes).
I run your script to extract reads mapping to two different chromosome(chrA_reads, chrB_reads), and detected the chimeric transcripts by find the overlaps. And in the same time I apply an independent methods to achieve that goal. I run the blastn and find the chimeric reads/trasncripts mapping to two different chromosome(known with fusion). Surprisingly, I found only 30% consistency between 2 methods. Specifically, about 70% cells don't detected by your method. For example, a read with CIGAR value 45S105M can map to the 2 fusion gene according to the blastn program, but haven't detected by your method. Actually, this kind of read can be extract in chrA_reads, but won't show in the chrB_reads.
**I was wondering that this might be related to the bam(output from cellrange) where one reads only has one record and without any tag saving the soft clip mapping information？**Could you help me find the answer?

Best wishes,
Shuting Jiang