Go Back   SEQanswers > Bioinformatics > Bioinformatics

Similar Threads
Thread Thread Starter Forum Replies Last Post
Interpretation of interaction term in DESeq2 fanli Bioinformatics 3 08-23-2016 02:53 PM
Differential expression of small subset of genes from RNASeq dataset Tryptamine RNA Sequencing 0 03-09-2016 10:17 AM
DESEQ2: How to use dispersion estimated from one dataset in another dataset EVELE Bioinformatics 3 07-01-2015 09:09 AM
DESeq2- How to get all levels of the interaction term Tom2013 Bioinformatics 1 03-20-2015 11:23 AM
DESeq2-test a subset of genes quinne5 RNA Sequencing 2 10-16-2013 06:19 AM

Thread Tools
Old 05-15-2020, 05:11 PM   #1
Junior Member
Location: Research Triangle

Join Date: Mar 2020
Posts: 1
Question DESeq2 interaction term results differ between full dataset & subset

Hi all, hoping someone has some insight on this because I am completely stumped...

I'm using DESeq2 for differential analysis between 2 different experiment/sequencing types (wholly different methods, not like SE vs PE) and with multiple treatment conditions vs. one control condition. I do have a batch effect so I have to include that term in my design as well.
ID Batch Condition SeqType
1 1 ctrl X
2 1 A X
3 1 B X
4 1 C X
5 2 ctrl X
6 2 A X
7 2 B X
8 2 C X
9 1 ctrl Y
10 1 A Y
11 1 B Y
12 1 C Y
13 2 ctrl Y
14 2 A Y
15 2 B Y
16 2 C Y

So the design ends up being " ~ Batch + SeqType + Condition + Condition:SeqType "

With reference levels for Condition=ctrl & SeqType=X, this gives me resultsNames of:
  • Intercept
  • Batch_2_vs_1
  • SeqType_Y_vs_X
  • Condition_A_vs_ctrl
  • Condition_B_vs_ctrl
  • Condition_C_vs_ctrl
  • SeqTypeY.ConditionA
  • SeqTypeY.ConditionB
  • SeqTypeY.ConditionC

It's my understanding that using results(dds, name="SeqTypeY.ConditionA") should give me the genes for which changes (ie. A vs. ctrl) in sequencing type Y are independent of changes in sequencing type X.

This analysis works and I get reasonable results for all of my conditions when I do not subset the data (i.e. conditions A, B, C, and ctrl are all in the same counts file input to the DESeqDataSet object, then results are called separately for each). However, when I subset by condition (i.e. input condition A & ctrl only), all the genes have padj>0.95 in the interaction term results.

Does anyone know why this is happening or if it matters? I'm concerned it means I'm doing something wrong in the multiple conditions analysis?

A related question - does the padj value correct for multiple comparisons between all genes within a specified results table (i.e. for all genes in the "SeqTypeY.ConditionA" results) or does it correct for multiple comparisons between all genes across all possible results tables (i.e. SeqTypeY.ConditionA vs. SeqTypeY.ConditionB vs. SeqTypeY.ConditionC)? I did the subset analysis because I was trying to answer this question for myself and I'm afraid I've just made my confusion worse....
cjag is offline   Reply With Quote

deseq2, differential expression, interactions, padjj

Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

All times are GMT -8. The time now is 09:27 AM.

Powered by vBulletin® Version 3.8.9
Copyright ©2000 - 2021, vBulletin Solutions, Inc.
Single Sign On provided by vBSSO