cuffcompare output statistic

[Yt Shen]

Dear sterding and yasu,

Did any of you get the information about the topic you talk about above? I also confused about it.

[sterding]

I have the same question. The 'base-level' Sn/Sp is reasonablly high, while the exon-level Sn/Sp is very low. Why like this?

[yasu]

I got cuffcompare output something like below:

==================
# Query mRNAs : 250371 in 250079 loci (24778 multi-exon transcripts)
# (282 multi-transcript loci, ~1.0 transcripts per locus)
# Reference mRNAs : 39609 in 25783 loci (31869 multi-exon)
# Corresponding super-loci: 14405
#--------------------| Sn | Sp | fSn | fSp
Base level: 53.1 38.8 - -
Exon level: 0.7 0.5 20.3 15.7
Intron level: 0.0 0.0 12.6 74.2
Intron chain level: 0.0 0.0 0.2 0.3
Transcript level: 0.0 0.0 0.0 0.0
Locus level: 0.0 0.0 0.3 0.0
Missed exons: 97628/219296 ( 44.5%)
Wrong exons: 164116/283022 ( 58.0%)
Missed introns: 163909/193288 ( 84.8%)
Wrong introns: 5780/32915 ( 17.6%)
Missed loci: 10965/25783 ( 42.5%)
Wrong loci: 154875/250079 ( 61.9%)

Total union super-loci across all datasets: 170453
(0 multi-transcript, ~0.0 transcripts per locus)
==================

I'd like to know the meaning of each statistic. Specifically, can I think that this data covers 14818 loci (calculated from 25783-10965 as shown in "Missed loci") at least one read?

I tried to look around related thread, but I cannot find out. Sorry for redundant thread if there are already related ones.

Thanks in advance.

-Yasu