STAR+HTSeq gene counts vs STAR+RSEM genes and how to combine isoforms to get the gene counts?

Question

[jennygenome]

Hi All,

• In our dataset, we observe the STAR+HTSeq counts are almost double for everygene compared to STAR+RSEM counts. Is there an explanation for this and how to interpret it?

ENSG00000179934.6 HTSeq_Counts    RSEM_expected_count Ratio(HTSeq/RSEM)
SAMPLE1   3350    1849    1.812
SAMPLE2   2668    1535    1.738
SAMPLE3   36  19  1.895
SAMPLE4   2647    1423    1.860
SAMPLE5   3120    1811    1.723
SAMPLE6   3234    1775    1.822
SAMPLE7   677 364 1.860
SAMPLE8   589 329 1.790
SAMPLE9   573 270 2.122
SAMPLE10  662 352 1.881
  

[swbarnes2]

You are probably going to have to put the command lines you used.

Answer 1

[swbarnes2]

One possible explanation...check the Aligned.ToTranscriptome.out.bam file. RSEM uses that, and reads that STAR can align and call with htseq-count might not be making it into the transcriptome file. I've seen that myself (in my case, the library prep had artifacts that STAR's aligner didn't mind, but STAR would not put those reads into the transcriptome file, so RSEM could not see them)