I have an mRNA-Seq experiment with a number of samples. After running them through tophat and cuffcompare I then proceeded to run the data through cuffdiff. I wasn't particularly interested in any specific differential expression testing, I just wanted to re-estimate gene level relative expression. I ran cuffdiff providing a curated combined.gtf generated by cuffcompare and four accepted_hits.sam files. According to the cufflinks manual the genes.fpkm_tracking file should contain 4 sets of qX_FPKM, qX_conf_lo and qX_conf_hi data. I was expecting q0 - q3 which is what I see in the isoforms, tss_group tracking files, but in the genes tracking file the first row (headers) lists 12 sets, q0 - q11. What's more, not all rows (loci) have the same number of data points. Many have 12 sets, consistent with the header row, some have only 4 sets which is the expected number and I've even found some rows with as many as 36 sets of FPKM and conf values.
I am using version 0.8.2 of cufflinks. This result occurs with both the prebuilt x86_64 binaries and binaries I compiled myself.
Has anyone else observed this behavior when providing more than two samples to cuffdiff?
Update:
I have now observed this behavior when only providing 2 input sam files to cuffdiff. The genes.fpkm_tracking file has q0 - q5 data sets.
I am using version 0.8.2 of cufflinks. This result occurs with both the prebuilt x86_64 binaries and binaries I compiled myself.
Has anyone else observed this behavior when providing more than two samples to cuffdiff?
Update:
I have now observed this behavior when only providing 2 input sam files to cuffdiff. The genes.fpkm_tracking file has q0 - q5 data sets.