Tophat / output without XS

Question

[Latifa]

Hello, I run my libraries with bowtie2 and tophat 2.1.2 to align my reads to a reference genome. When I looked my output accepted_hits.bam I realized that I don't have the XS tag where the strand should be specified. My result looks like it:

HWI-ST957:98:D10H2ACXX:8:1308:9210:179192 256 ChrUn 1610837 0 41M * 0 0 TGTCCCGAGGGACGGAGGAGGCTAGGTTAGCCGAAAGATGG @@8BDDDDH<fff?<?<?ffegggec ?f8="CAHFIAHD">A AS:i:-5 XN:i:0 XM:i:1 XO:i:0 XG:i:0 NM:i:1 MD:Z:0A40 YT:Z:UU NH:i:10 CC:Z:= CP:i:86011202 HI:i:0

Do someone got something similar? I am doing it because I want to run cufflinks to identify new isoforms, then run RSEM, and for RSEM I need the XS tag.

[piyushjo]

Is this RNA-seq data? Any particular reason you are using tophat and not new aligners like HISAT2 or STAR? In HISAT2 you have to specify strandness, in default condition it assumes unstranded, could be the same issue with tophat.

Answer 1

[i.sudbery]

Tophat is deprecated software. Unless you have a good reason (such as replicating an old analysis) to use Tophat over more modern, accurate and actively supported software, such as STAR or HISAT2.

Likewise Cufflinks is superseded by StringTie, although there hasn't actaully been a deprecation notice.

In this case, you are not getting an XS tag because the read is not spliced. In unstranded RNAseq data the aligner can only assign a strand when the read maps across an intron and it can use the orientation of the splice-site motif to assign strand.