rsem calculate expression using STAR. Reference genome format?

Question

[dtatarak]

I am using rsem to calculate expression levels in a dataset of paired end reads. I want to use STAR as the aligner. I have written the command like this:

rsem-calculate-expression --star -p 16 --paired-end DT_01_R1.fastq DT_01_R2.fastq /path/to/reference/genome/ rsem_output/DT_01

I'm pretty sure this will work, What I don't know is this: do I need to generate an annotated reference genome for STAR before running this? I'm using the zebrafish genome, and I have downloaded both the .fasta file and the .gtf file from Ensembl. What do I need to do with these before I run the aligning and expression calculation I've written above? Or do I just point it to the directory containing the .fasta and .gtf files? Thanks very much!

[dtatarak]

Ok great thanks! I thought that was the case, but the manual didn't make it clear to me.

Answer 1

[h.mon]

You have to prepare the transcriptome reference first, see rsem-prepare-reference. See a tutorial here.

Alternatively, you can map with STAR to the genome (you will have to prepare the index yourself), using STAR --quantMode transcriptomeSAM, and then use STAR's Aligned.toTranscriptome.out.bam output as input to rsem-calculate-expression.

[k.kathirvel93]

Hi h.mon How we can do quantification with RSEM by using STAR aligned .bam(aligned sepeartely by using STAR) file. Can i get the code?

[h.mon]

What is the command you use to align with STAR?

[k.kathirvel93]

STAR --genomeDir /home/STAR/Indexed_Reference/ --runThreadN 16 \
    --sjdbGTFfile /home/ensemble_GRCh38.77/Homo_sapiens.GRCh38.77.gtf \
    --readFilesIn /home/_1.fastq  /home/2.fastq --sjdbOverhang 101 \
    --outSAMtype BAM SortedByCoordinate \
    --outFileNamePrefix /home/STAR_Output/Mapping/output

Should I mention --quantMode also?

[h.mon]

Yes, you should have used --quantMode TranscriptomeSAM --quantTranscriptomeBam IndelSoftclipSingleend.

However, RSEM is finicky with the annotation, I would suggest you follow its tutorial because you may end up with lots of bam files that won't work with RSEM, unless the annotation strictly follows what RSEM expects.