Empty FPKM estimates after cuffcompare

[henning]

I ran cufflinks on several BAM files and get tracking.gtf files looking like (truncated lines):

Chr1    Cufflinks   transcript  22746   23685   1000    +   .   gene_id "CUFF.8"; transcript_id "CUFF.8.1"; FPKM "32.3697173163"; frac "1.000000"; conf_lo "28.299584"; conf_hi "36.439851"; cov "52.362831";
Chr1    Cufflinks   exon    22746   23020   1000    +   .   gene_id "CUFF.8"; transcript_id "CUFF.8.1"; exon_number "1"; FPKM "32.3697173163"; frac "1.000000"; conf_lo "28.299584"; conf_hi "36.439851"; cov "$
Chr1    Cufflinks   exon    23096   23685   1000    +   .   gene_id "CUFF.8"; transcript_id "CUFF.8.1"; exon_number "2"; FPKM "32.3697173163"; frac "1.000000"; conf_lo "28.299584"; conf_hi "36.439851"; cov "$
Chr1    Cufflinks   transcript  79944   80186   1000    +   .   gene_id "CUFF.3"; transcript_id "CUFF.3.1"; FPKM "3077.9273423595"; frac "1.000000"; conf_lo "1765.494714"; conf_hi "4390.359971"; cov "2714.84$
Ch

and then run cuffcompare to compare with a merged gtf file as

cuffcompare -o myprefix -r merged.gtf transcripts.gtf

but, the myprefix.tracking only gives me FPKM estimates equal zero for all genes:

TCONS_00000001  XLOC_000001 XLOC_000001|TCONS_00000001  c   q1:CUFF.8|CUFF.8.1|100|0.000000|0.000000|0.000000|0.000000|-
TCONS_00000002  XLOC_000002 XLOC_000002|TCONS_00000002  =   q1:CUFF.11|CUFF.11.1|100|0.000000|0.000000|0.000000|0.000000|-
TCONS_00000003  XLOC_000003 XLOC_000003|TCONS_00000003  =   q1:CUFF.2|CUFF.2.1|100|0.000000|0.000000|0.000000|0.000000|-
TCONS_00000004  XLOC_000004 XLOC_000004|TCONS_00000004  =   q1:CUFF.4|CUFF.4.1|100|0.000000|0.000000|0.000000|0.000000|-
TCONS_00000005  XLOC_000005 XLOC_000005|TCONS_00000006  c   q1:CUFF.16|CUFF.16.1|100|0.000000|0.000000|0.000000|0.000000|-
T

note that CUFF.8 in the transcripts.gtf has a non-zero FPKM estimate but this does not propagate to the myprefix.tracking file.

Any ideas why? Grateful for answers.

[bgibb]

Have you tried running it with Cufflinks-1.1.0? (released last week)

I noticed a similar problem with the .tmap files generated by cuffcompare in release 1.0.3, however that problem seems to have been corrected in the most recent release.

[henning]

Thanks for the tip, I tried Cufflinks-1.1.0 and indeed this solves the problem!