strange output from RSEM

[tujchl]

hi,
I use RSEM (http://deweylab.biostat.wisc.edu/rsem/README.html) to analysis my RNA-seq data and want to summarize expression level of each RefSeq ID.
the command I used is as below:
rsem-calculate-expression -p 7 --fragment-length-min 51 --fragment-length-max 51 --fragment-length-mean 51 --fragment-length-sd 0
BTW, the length of all my reads is 51bp

I found that many isoform from the same gene got quite different TPM and FPKM. An example is as below:
transcript_id gene_id length effective_length expected_count TPM FPKM IsoPct
NM_001130147.1 CCDC77 2359 2309.00 317.27 16.83 11.21 100.00
NM_001130148.1 CCDC77 2305 2255.00 2.23 0.12 0.08 100.00

as you see. NM_001130147.1 and NM_001130148.1 are different isoform of CCDC77 and the mRNA seq is quite similar with each other. but the TPM as well as FPKM is quite different which I regard it as err.

Does anyone meet the err? can you help me?

[dpryan]

Why do you consider that to be an error? Typically there's only one major isoform.

[tujchl]

thank you for your replying, dpryan:

because the mRNA seq of these two isoforms are nearly the same. Most of the reads mapped to NM_001130147.1 can also be mapped to NM_001130148.1. So I think the expression level of these two isoforms should be nearly the same.

Am I right?

[dpryan]

No, just because most of the reads can map to both doesn't mean the expression was more or less the same. That's the entire point of programs like RSEM, have a read through the paper for details.

[tujchl]

Thank you, dpryan
To be honest, I didn`t read the papper. Now I think it`s the time to read it.