Partially mapped reads using STAR

[Stuti Agrawal]

Hi,

I am trying to quantify the expression for some RNA-seq datasets using STAR and RSEM. I performed the following steps:

(a) Prepared the reference using rsem-prepare-reference. I provided the GTF file in this step.

(b) Used the ref_name.idx.fa file generate to create the genome index for STAR

(c) Aligned the RNA-seq reads to the index thus generated using star-2-pass. I used the parameters to prevent indels and perform end to end matching as mentioned in previous posts.

(d) Ran convert-sam-for-rsem on the resulting alignment file.

However, when I do that, I get the error that "Number of first and second mates in read ABC are not matched!". When I check my BAM file for the alignment of the two mates of this read, I find that there is one alignment for the first mate and two alignments for the second mate. 

Is this something that RSEM can handle? If so, how should I proceed? 

If not, is there a way to produce equal number of partially mapped reads using STAR?

Thanks,

Stuti

[Alexander Dobin]

Hi Stuti,

please send me the Log.out file, as well as the lines from the BAM files.

Note, that instead of mapping to the transcriptome, you could map to the genome (i.e. do not use RSEM reference for generating STAR genome, use genome FASTA files and the GTF file), and then use --quantMode TranscriptomeSAM to generate BAM alignments comaptible with RSEM.

Cheers

Alex