[erikjlar]

Thanks again for the input. I don't want to bother Cole with more questions - but maybe someone else can comment on alternative solutions if one just wants to do a simple gene-level summarization of reads.

It seems to me that for this scenario, the sophistication of cufflinks can get in the way of getting a robust expression estimate (see my previous post and picture to see what I mean).

Basically, for the example above I would be happy to just get FPKM values, or even read counts, based on all MYH11 exons - but excluding positions that ambiguously overlap with other genes.

Someone must be doing this sort of "array-style" abundance estimates?

[dglemay]

gene-level estimates

You could use the genes.fpkm_tracking files to get the FPKMs summarized at the gene level.

[Jon_Keats]

Update

I've had a chance to check my original dataset for the frequency of "FAIL" genes now with the newest version of Cufflinks. It is much improved with the number of FAIL genes now being between 73-238 compared to the previous version where the "FAIL" genes were between 8000-12,000 genes. There is still a trend for more "FAIL" genes as the number of aligned reads increases which I find odd but at this frequency it seems acceptable.