gffread converts stop codons into CDS

[jg3197]

Hi all,

I am having a problem with duplicate values in my gtf cufflinks input file. I used gffread with the -T option, but looking at my output I get strange results. Here is an example input gene record:

transcript 438613 486325 0.59 - .
tts 438613 438613 . - .
exon 438613 438662 . - .
stop_codon 438659 438661 . - 0
intron 438663 438839 0.71 - .
intron 439022 485468 1 - .
intron 485546 486140 0.83 - .
CDS 438659 438662 0.71 - 1
CDS 438840 439021 1 - 0
exon 438840 439021 . - .
CDS 485469 485545 0.99 - 2
exon 485469 485545 . - .
CDS 486141 486147 0.83 - 0
exon 486141 486325 . - .
start_codon 486145 486147 . - 0
tss 486325 486325 . - .

and the corresponding output:

exon 438613 438662 . - .
CDS 438659 438662 0.71 - 1
CDS 438661 438661 . - 0
exon 438840 439021 . - .
CDS 438840 439021 1.00 - 0
exon 485469 485545 . - .
CDS 485469 485545 0.99 - 2
CDS 486141 486147 0.83 - 0
exon 486141 486325 . - .
CDS 486147 486147 . - 0

I can see that line 3 of the output, the CDS start and stop position corresponds to the last base of the stop codon for the first exon in the input file (line 4), and the last line of the output corresponds to the last base of the start codon (input file line 15). Is this normal behaviour? If so, can somebody explain to me why it does this, and if not, whether the file is suitable for downstream analysis? Would it be better instead to grep the exon and cds fields from the input file for use in cufflinks analysis?

thank you in advance,

J.