RNA seq: tuxedo protocol - separate non-overlapping genes of same XLOC id in cuffdiff output

Question

[CHANG]

I notice that there're two non-overlapping genes (PHLDB3, ETHE1) , but they have the same XLOC id = XLOC_018276

cuffmerge's output - merged.gtf

gene_id "XLOC_018276"; transcript_id "TCONS_00085241"; exon_number "16"; gene_name "PHLDB3"; oId "NM_198850"; nearest_ref "NM_198850"; class_code "="; tss_id "TSS34843"; p_id "P15382";
chr19 Cufflinks
exon 44009900
44009902 .
- .
gene_id "XLOC_018276"; transcript_id "TCONS_00085242"; exon_number "1"; gene_name "ETHE1"; oId "CUFF.30707.12"; nearest_ref "NM_014297"; class_code "j"; tss_id "TSS34842";
chr19 Cufflinks
exon 44010944
44011054 .

Thus, when I run cuffdiff, I get one value of fold change for both genes.

test_id    gene_id    gene    locus    sample_1    sample_2    status    value_1    value_2    log2(fold_change of value2/value1)    test_stat    p_value    q_value    significant
XLOC_018276    XLOC_018276    ETHE1,PHLDB3    chr19:43978502-44031860    tumor    normal    OK    222.08    450.931    1.02183    1.29159    0.00005    0.00321114    yes

How do I get separate fold changes for ETHE1 and PHLDB3?

[Devon Ryan]

If you look at your alignments, do you see a lot of reads mapping in between the two genes? Cufflinks will merge them in that case.

[CHANG]

There are more reads in ETHE1 than PHLDB3 , it doesn't seem like cufflink should merge them. I have attached a screen shot of igv

[Devon Ryan]

N.B., I updated your post so that the image shows.

My guess is that cufflinks sees the reads between the two genes as joining them. They may in fact be transcribed as a unit sometimes, so cufflinks might not be wrong in doing that.