STAR mapping and HTSeq for stranded RNA-seq

[Maoting Chen]

Hi,

I used Illumina Stranded mRNA Prep Ligation for my RNA-seq library prep. 

According to the STAR manual, it seems the default parameters are good with stranded RNA-seq mapping. But does it influence the results if I add the --outSAMstrandField intronMotif (recommended for unstranded data) when mapping my stranded RNA-seq?

 

I also plan to use HTSeq to do the expression analysis. For stranded RNA-seq, which parameters should I use?

Thanks,

Maoting

[Alexander Dobin]

Hi Maoting,

--outSAMstrandField intronMotif will influence the results in the following way: it will remove all alignments that contain only non-canonical junction(s).

HTSeq will work with STAR BAMs, no special parameters are required. You will need to specify strandedness as HTSeq parameter (I think it's -s).

Cheers

Alex

[Maoting Chen]

Hi Alex,

Thanks for your reply.

I have one more question about the two-pass mapping. 

I have 24 RNA-seq samples with 4 conditions from C.elegans, and 6 replicated for each condition. And I want to detect the alternative splicing events using these samples. I used the two-pass mapping in STAR. 

For the first round, I used the default setting and got the SJ.out.tab from each sample.

For the second round, I built the genome index for each sample separately:

STAR --runMode genomeGenerate --genomeDir genome_index2  --genomeFastaFiles Fasta_files  --sjdbFileChrStartEnd SJ.out.tab  --sjdbGTFfile .gtf  --runThreadN 1

Then I did the second mapping using the sample-specific genome index2.

However, in the manual it's recommended to concatenate the SJ.out.tab files from all samples together and build the same genome index for all the samples.

I am wondering if this difference will influence a lot? Do you recommend me to redo the second mapping as suggested in the manual?

Thanks,

Maoting

[Alexander Dobin]

Hi Maoting,

sorry for the belayed reply.

Concatenating junctions from all samples will increase sensitivity for low-expressed junctions. However, if you have a lot of samples, and a lot of novel junctions, it may result in a too large search space and slow down the mapping and reduce the % of uniquely mapped reads -  so the per-sample 2nd pass is a "safer" option.

Cheers

Alex