[metheuse]

Hav you checked your alignments with a genome browser? Load the SAM file and the GFF file produced by dexseq_prepare in, e.g., IGV, and look at one of the loci with zero counts. If there really are no reads, you experiment has failed (or you are using a wrong annotation file).

I converted the bam file to bed and intersected it with "chr1 29385323 29385364 ENSG00000159023:023" (which has zero count in the dexseq_count.py output)
This resulted in 71 intersecting reads. Here are the first 10 of them:

chr1    29379725        29391495        HWI-ST1235:101:C1WW9ACXX:6:1312:1770:71045/1    50      +
chr1    29379725        29391495        HWI-ST1235:101:C1WW9ACXX:6:2116:17025:56846/2   50      -
chr1    29379731        29391501        HWI-ST1235:101:C1WW9ACXX:6:2307:18153:8715/1    50      +
chr1    29379734        29391504        HWI-ST1235:101:C1WW9ACXX:6:2314:2163:52123/2    50      +
chr1    29379736        29391506        HWI-ST1235:101:C1WW9ACXX:6:2314:2163:52123/1    50      -
chr1    29379741        29391511        HWI-ST1235:101:C1WW9ACXX:6:2313:2799:76858/1    50      +
chr1    29379742        29391512        HWI-ST1235:101:C1WW9ACXX:6:2210:9177:71165/1    50      +
chr1    29379742        29391512        HWI-ST1235:101:C1WW9ACXX:6:1101:15865:15952/2   50      -
chr1    29379742        29391512        HWI-ST1235:101:C1WW9ACXX:6:1307:5858:86759/2    50      -
chr1    29379742        29391512        HWI-ST1235:101:C1WW9ACXX:6:1308:20389:32047/1   50      -

This should mean both my reads and the annotation file has no problem.

[metheuse]

By the way, these are the commands I used:

samtools index 21722_mapped_hg19/accepted_hits.bam
samtools view 21722_mapped_hg19/accepted_hits.bam >21722_accepted_hits.sam
sort -k1,1 -k2,2n 21722_accepted_hits.sam >21722_accepted_hits_sorted.sam
python ~/scripts_64/dexseq_count.py -p yes -s no ~/scripts_64/Homo_sapiens.GRCh37.71.DEXSeq.chr.gff 21722_accepted_hits_sorted.sam KARPAS299_CEP.txt

[metheuse]

I just noticed the chromosome name of the gff file doesn't contain "chr":

1    Homo_sapiens.GRCh37.71.gtf      exonic_part     11869   11871   .       +       .       transcripts "ENST00000456328"; exonic_part_number "001"; gene_id "ENSG00000223972"

This is probably the reason. I've added chr to each line and see if it works.

[sindrle]

I also have a question about warnings and dispersion:

ecs <- estimateSizeFactors( ecs )
the matrix is either rank-deficient or indefinite

ecs <- fitDispersionFunction( ecs )
Too much damping - convergence tolerance not achievable



And does this look ok?

Thanks! First time DEXSeq..

[czelin]

Dear all,

I have recently started with exon-wise analysis and would appreciate your help.

I have paired 100bp reads. I have prepared the annotation file with DEXSeq python scripts. What I have realized is that when I have a shorter exon (e.g. 150bp) the number of tags is 0. In IGV I can see lots (>500) of reads spanning this exon. Somehow few of the reads were counted for the control samples and none for comparison and this exon was reported as significantly DE.
I suspect that is because the two read pairs are longer than this exon and they overlapped adjacent exon[s]. This caused the reads be considered as "_ambiguous_readpair_position". If my understanding is correct, is there any way to solve the short-exon issue? If my understanding is wrong, could you please correct me?

[areyes]

Hi czelin,

This is unlikely, since reads that overlap to many exons should be counted once per each exon. You could check whether the reads that are ignored by the script are properly paired or if they are mapping to multiple regions on the genome?

Alejhandro