RSEM and STAR

Question

[rajdeepboral00]

I have got a gene counts matrix using RSEM and STAR both but there is a high discrepancy between the gene numbers like for gene x, there are 0 for star(used --quantmode GeneCounts) and for RSEM the expected count is 900, so what will i trust?

Answer 1

[swbarnes2]

STAR can't tolerate reads that have ambiguous alignment positions, RSEM can model how to count them. So see if the reads for that gene have low mapping score.

If RSEM is doing the aligning too, it might be using different STAR parameters than you used with STAR.
  In general, I'd trust RSEM, its algorithm is smarter than STAR.

[rajdeepboral00]

Actually for analysing with RSEM i had firstly aligned with star and used --quantmode TranscriptomicSam GeneCounts and then with the gene level counting i did the RSEM quantification. I need the isoform level quantification too actually