RSEM output interpretation

Question

[komal.rathi]

Hi everyone,

I am using RSEM to get FPKM values as a measure of gene quantification. I am having trouble understanding a couple of things in the output files:

In rsem.genes.results:

gene_id transcript_id(s)    length  effective_length    expected_count  TPM FPKM    id
ENSG00000206588.1   ENST00000383861.1   164 17.24   0   0.01    0.00    sample1
ENSG00000206652.1   ENST00000383925.1   164 17.24   0   0.01    0.00    sample2
ENSG00000206737.1   ENST00000384010.1   164 17.24   0   0.01    0.00    sample3
ENSG00000207513.1   ENST00000384782.1   164 17.24   0   0.01    0.00    sample4
ENSG00000207067.1   ENST00000384339.1   132 04.81   0   0.01    0.01    sample5
ENSG00000272220.1   ENST00000607347.1   118 02.27   0   0.01    0.01    sample6
ENSG00000275254.1   ENST00000610451.1   147 08.54   0   0.01    0.01    sample7
ENSG00000275994.1   ENST00000384096.1   131 04.37   0   0.01    0.01    sample8
ENSG00000280498.1   ENST00000628458.1   135 04.19   0   0.01    0.01    sample9
ENSG00000280547.1   ENST00000630628.1   138 07.08   0   0.01    0.01    sample10
ENSG00000207051.1   ENST00000384323.1   126 03.33   0   0.03    0.02    sample11

Why is the expected_count sometimes 0 but the corresponding TPM and FPKM may or may not be 0?

And in rsem.isoforms.results:

transcript_id   gene_id length  effective_length    expected_count  TPM FPKM    IsoPct  id
ENST00000494424.1   ENSG00000000003.14  820 624.19  0   0   0   0   sample1
ENST00000494424.1   ENSG00000000003.14  820 614.82  0   0   0   0   sample2
ENST00000494424.1   ENSG00000000003.14  820 632.04  0   0   0   0   sample3
ENST00000494424.1   ENSG00000000003.14  820 604.99  0   0   0   0   sample4
ENST00000494424.1   ENSG00000000003.14  820 621.9   0   0   0   0   sample5
ENST00000494424.1   ENSG00000000003.14  820 621.33  0   0   0   0   sample6
ENST00000494424.1   ENSG00000000003.14  820 650.97  0   0   0   0   sample7
ENST00000494424.1   ENSG00000000003.14  820 640.21  0   0   0   0   sample8

The length for a single transcript_id remains constant per sample but why does the effective_length changes?

Thanks!

[glady]

It Shouldn't be like this. Probably there is some problem with the parameters you are using.

[komal.rathi]

I am using the following parameters:

rsem-calculate-expression \
--paired-end \
--no-bam-output \
--quiet \
--no-qualities \
-p {threads} \
--forward-prob 0.5 \
--seed-length 25 \
--fragment-length-mean -1.0 \
--bam {bam} {genomedir} {prefix}

[glady]

probably its a bug, I don't know the exact reason why, but its an error. I used RSEM a number of times, but there was never such a situation.