why does ScanNeo realign with bwa?

[indapa]

Really interesting paper and tool. I'm wondering why reads are converted to single end and then realigned with bwa, as described in the publication?

Retained reads were converted to single-end reads in FASTQ file using BEDTools (Quinlan and Hall, 2010) and re-aligned using BWA-MEM (Li, 2013) which reports chimeric alignments with an ‘SA’ tag in the alignment records. Finally, ScanNeo applied transIndel (Yang et al., 2018) to determine small to large-sized indels from the RNA-Seq reads. Indel calls were reported in variant call format (VCF) file.

[dolittle007]

Because in the previous step, we removed spliced reads with inferred transcriptional directions but not carrying indels within their alignment. it made paired-end reads mapping impossible. Additionally, transIndel need chimeric alignments from BWA-MEM to infer small to large-sized Indels. Thus, BWA-MEM was applied to single-end reads mapping.