DESeq2 adjusted p-values of 0 when p-value > 0.05

Question

[Adrian]

DESeq2 assigns adjusted p-values of 0 in some cases when the p-value is above 0.05:

name    baseMean          log2FoldChange   lfcSE             stat              pvalue              padj  gene_description
BCL6B   1.16290763279826  4.8304810831086  2.84617260952113  1.69718486747764  0.0896617151432926  0     BCL6B_transcription_repressor
C4B     1.16290763279826  4.8304810831086  2.84617260952113  1.69718486747764  0.0896617151432926  0     complement_C4B_(Chido_blood_group)
GATD3B  1.16290763279826  4.8304810831086  2.84617260952113  1.69718486747764  0.0896617151432926  0     glutamine_amidotransferase_like_class_1_domain_containing_3B
IL12A   1.16290763279826  4.8304810831086  2.84617260952113  1.69718486747764  0.0896617151432926  0     interleukin_12A
KCNJ13  1.16290763279826  4.8304810831086  2.84617260952113  1.69718486747764  0.0896617151432926  0     potassium_inwardly_rectifying_channel_subfamily_J_member_13
MRC1    1.16290763279826  4.8304810831086  2.84617260952113  1.69718486747764  0.0896617151432926  0     mannose_receptor_C-type_1
PNLDC1  1.16290763279826  4.8304810831086  2.84617260952113  1.69718486747764  0.0896617151432926  0     PARN_like,_ribonuclease_domain_containing_1
SGSM3   1.16290763279826  4.8304810831086  2.84617260952113  1.69718486747764  0.0896617151432926  0     small_G_protein_signaling_modulator_3
TRPV2   1.16290763279826  4.8304810831086  2.84617260952113  1.69718486747764  0.0896617151432926  0     transient_receptor_potential_cation_channel_subfamily_V_member_2

Coincidentally these genes aren't really expressed in my sample.

If I sort to filter based on both pvalue and padj then the results make sense.

Does anyone know why the weird behavior of padj being 0?

Thanks!

[Carlo Yague]

Is it normal that all your genes have EXACTLY the same baseMean, log2FoldChange and pvalue ? Something smells wrong here...

[Ram]

It could be a merge operation gone wrong - DESeq2 does not generate the gene_description column.

[ATpoint]

Can you show some code as this is not default DESeq2 output, also sessionInfo() please.

[Adrian]

Thanks for catching that, I looked through the code again and there was a step when gene_Descriptions are added where NA padj values are converted to 0.

[ATpoint]

Without code I (or others) cannot help. As Carlo and Ram mention, this is not DESeq2 output and all values are the same, therefore suspicious of merging or joining gone wrong.

[h.mon]

Did you check the distribution of pvalues (before adjustment)? I don't recall ever seeing an adjusted p-value being smaller than the unadjusted p-value.

[h.mon]

Of interest:

Q-Value Less than P-Value

[rodolfo.peacewalker]

Hi Adrian!

Coincidentally these genes aren't really expressed in my sample

I'm not sure whether this is the cause, did you pre-filter the genes according to their abundance (counts)? I mean, removing those showing low counts. Please, would you like to show your code?

Best regards!