How Can Cufflinks Assemble De Novo Transcripts With Many Exons When They Have 0 Fpkm

Question

[biorepine]

Hi, I have seen many studies where they assembled transcriptomes (RNA-Seq) using cufflinks and usually it finds many novel transcripts but surprisingly most of them have 0 FPKM expression levels. Has any one noticed this ?

A good ex:

chr5    Cufflinks    transcript    19702133    19761803    1    -    .    gene_id "XLOC_150252"; transcript_id "TCONS_00425102"; FPKM "0.0000000000"; frac "0.000000"; conf_lo "0.000000"; conf_hi "0.000000"; cov "0.000000";
chr5    Cufflinks    exon    19702133    19702839    1    -    .    gene_id "XLOC_150252"; transcript_id "TCONS_00425102"; exon_number "1"; FPKM "0.0000000000"; frac "0.000000"; conf_lo "0.000000"; conf_hi "0.000000"; cov "0.000000";
chr5    Cufflinks    exon    19704533    19704567    1    -    .    gene_id "XLOC_150252"; transcript_id "TCONS_00425102"; exon_number "2"; FPKM "0.0000000000"; frac "0.000000"; conf_lo "0.000000"; conf_hi "0.000000"; cov "0.000000";
chr5    Cufflinks    exon    19705319    19705437    1    -    .    gene_id "XLOC_150252"; transcript_id "TCONS_00425102"; exon_number "3"; FPKM "0.0000000000"; frac "0.000000"; conf_lo "0.000000"; conf_hi "0.000000"; cov "0.000000";
chr5    Cufflinks    exon    19706257    19706391    1    -    .    gene_id "XLOC_150252"; transcript_id "TCONS_00425102"; exon_number "4"; FPKM "0.0000000000"; frac "0.000000"; conf_lo "0.000000"; conf_hi "0.000000"; cov "0.000000";
chr5    Cufflinks    exon    19706708    19706805    1    -    .    gene_id "XLOC_150252"; transcript_id "TCONS_00425102"; exon_number "5"; FPKM "0.0000000000"; frac "0.000000"; conf_lo "0.000000"; conf_hi "0.000000"; cov "0.000000";
chr5    Cufflinks    exon    19708313    19708453    1    -    .    gene_id "XLOC_150252"; transcript_id "TCONS_00425102"; exon_number "6"; FPKM "0.0000000000"; frac "0.000000"; conf_lo "0.000000"; conf_hi "0.000000"; cov "0.000000";
chr5    Cufflinks    exon    19709434    19709586    1    -    .    gene_id "XLOC_150252"; transcript_id "TCONS_00425102"; exon_number "7"; FPKM "0.0000000000"; frac "0.000000"; conf_lo "0.000000"; conf_hi "0.000000"; cov "0.000000";
chr5    Cufflinks    exon    19710326    19710331    1    -    .    gene_id "XLOC_150252"; transcript_id "TCONS_00425102"; exon_number "8"; FPKM "0.0000000000"; frac "0.000000"; conf_lo "0.000000"; conf_hi "0.000000"; cov "0.000000";
chr5    Cufflinks    exon    19712308    19712467    1    -    .    gene_id "XLOC_150252"; transcript_id "TCONS_00425102"; exon_number "9"; FPKM "0.0000000000"; frac "0.000000"; conf_lo "0.000000"; conf_hi "0.000000"; cov "0.000000";

.........

[David Fredman]

Which paper/dataset, and which versions of tophat and cufflinks was used for the above analyses? There were some bugs in various older builds leading to assignment of 0 FPKM in cufflinks and cuffdiff runs, not sure how many of them still persist in v2.2+.. see e.g.

• http://seqanswers.com/forums/showthread.php?t=15473
• http://seqanswers.com/forums/showthread.php?t=14864
• http://seqanswers.com/forums/showthread.php?t=8218

If there are alternate isoforms for the genes in question, it could potentially also be linked to cufflinks consideration of multiple isoforms per gene and probabilistic assignment of reads?

Due to similar issues and the somewhat black box nature of cufflinks (and limited design matrix setups), I've turned to other tools for abundance estimates (eXpress) and differential expression analyses (DESeq2, limma).

[pengchy]

Cufflinks is used to assembly transcripts, the FPKM was used to filter the results. For this purpose, cufflinks is worth to try.

[biorepine]

I posted few lines as biostar has limit on word size.

[bmpbowen]

I have noticed this as well in my RNAseq data have have yet to find an explanation.