Incosistent fold change signal : Differentially Expressed miRNA analysis : DEseq VS. Cuffdiff
Entering edit mode
6.0 years ago
illinois.ks ▴ 180

I am working with differentially expressed miRNAs from two conditions(control vs treated) Each condition has three replicates.

I have done this with three tools (edgeR, DESeq, Cuffdiff)
I knew that they can return different DEGs.. (Differentially Expressed miRNA analysis : DEseq VS. Cuffdiff)

However, what I found that the signal(positive/negative) of fold change is also not consistent (which doesn't make sense )

So, I am wondering whether I am making mistakes or not.

For EdgeR,DESeq, 

I have used bowtie2 for mapping to mmu10 mouse genome.
Then, samtools for sorted bam file and htseq-count for making countMatrix for each of 6 samples.
And concatenate them and done with DESeq/EdgeR

Here is the result.


        logFC    logCPM    LR    PValue    FDR
Mir143    -1.04130888    17.27205116    42.69394048    6.40E-11    3.36E-08
Mir320    1.965640966    8.975163956    28.9373577    7.48E-08    1.96E-05
Mir34a    1.685443326    12.74294283    20.2680951    6.73E-06    0.000883493
Mir32    -1.738969556    9.232269026    19.56020915    9.75E-06    0.001023523
Mir192    -0.879327264    15.34612311    16.94540076    3.85E-05    0.003366161

gene_id    gene    locus    sample_1    sample_2    status    value_1    value_2    log2(fold_change)    test_stat    p_value    q_value
XLOC_010256    Mir143    chr18:61639652-61665538    C    A    OK    10911600    61309100    2.49024    372.553    0.00005    0.00256667
XLOC_006383    Mir320    chr14:70443509-70443591    C    A    OK    92887.6    889374    3.25923    6.88674    0.0004    0.0125714
XLOC_014975    Mir34a    chr4:150068453-150068555    C    A    OK    1563360    6271810    2.00423    3.8403    0.00005    0.00256667
XLOC_015205    Mir32    chr4:56876012-56947429    C    A    OK    1220440    441142    -1.46809    -2.07333    0.00025    0.00875
XLOC_010429    Mir192    chr19:6264843-6264932    C    A    OK    115868    6810790    5.87727    20.3325    0.00005    0.00256667


X.Intercept. conditiontreated deviance converged pval padj
Mir143 16.9706648173215 -1.09288245394718 3.01272180511012 TRUE 0.000472698549358053 0.0620416846032444
Mir320 6.88785951514642 1.94583234864848 2.9487233864551 TRUE 1.91723318332393e-08 1.00654742124506e-05
Mir34a 10.907653786997 1.65520255918831 5.02147897477293 TRUE 3.57415829050911e-05 0.00938216551258642
Mir32 9.1281907527577 -1.78570498751835 2.76778136983314 TRUE 0.000402585086401896 0.0620416846032444
Mir192 14.9802788064507 -0.91961433768963 2.94113880567465 TRUE 0.00318173760664342 0.15185565849889

I guess that since DEseq and edgeR was referring the same count matrix created by htseq-count, they show similar results.
But Cuffdiff results looks very different.

Could you please somebody give comments for this? 

DEG; miRNAs edgeR; DESeq; Cuffdiff • 2.4k views
Entering edit mode

As far as I'm concerned cuffdiff is a black box...which makes me very hesitant to trust it. Having said that, nothing beats independent validation. Get some additional samples and do some qPCR or see if one of the results makes vastly more biological sense (given whatever is already known about your system from the literature).

Entering edit mode

Thank you so much for your comments! 



Login before adding your answer.

Traffic: 1348 users visited in the last hour
Help About
Access RSS

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.

Powered by the version 2.3.6