Question: DESeq2: normalization of experiments sequenced separately
gravatar for uguy
5 months ago by
uguy0 wrote:


I am analyzing RNAseq data of two experiments with multiple time points and their own control condition. These two experiments were sequenced in two separate runs. I wonder if I can merged the two raw counts tables and normalize them using the DESeq2 size factors method, or do I need to normalized each table separately with DESeq2 package and then use a quantile normalization on the two datasets ?

Thanks !

ADD COMMENTlink modified 5 months ago • written 5 months ago by uguy0

Thanks for answering me. There is two type of acclimated cells that were exposed to several stresses. The two types of acclimation were separated for sequencing step in two runs. The time points are not the same accross experiments and there is three biological replicates for each time point.

And yes, I want to analyze the normalized counts together but also compare the differential expression (log2FC) accross stress conditions and acclimation conditions.

Here the experiment design:

label reference acclimation_condition stress_condition time_point biological_replicate Run
acc1_Ct_T0_A None acc1 None(control) Ct A run1
acc1_Ct_T0_B None acc1 None(control) Ct B run1
acc1_Ct_T0_C None acc1 None(control) Ct C run1
acc1_Stress2_T0-5_A acc1_Ct_T0 acc1 Stress2 T0.5 A run1
acc1_Stress2_T0-5_B acc1_Ct_T0 acc1 Stress2 T0.5 B run1
acc1_Stress2_T0-5_C acc1_Ct_T0 acc1 Stress2 T0.5 C run1
acc1_Stress2_T1_A acc1_Ct_T0 acc1 Stress2 T1 A run1
acc1_Stress2_T1_B acc1_Ct_T0 acc1 Stress2 T1 B run1
acc1_Stress2_T1_C acc1_Ct_T0 acc1 Stress2 T1 C run1
acc1_Stress2_T2_A acc1_Ct_T0 acc1 Stress2 T2 A run1
acc1_Stress2_T2_B acc1_Ct_T0 acc1 Stress2 T2 B run1
acc1_Stress2_T2_C acc1_Ct_T0 acc1 Stress2 T2 C run1
acc1_Stress2_T3_A acc1_Ct_T0 acc1 Stress2 T3 A run1
acc1_Stress2_T3_B acc1_Ct_T0 acc1 Stress2 T3 B run1
acc1_Stress2_T3_C acc1_Ct_T0 acc1 Stress2 T3 C run1
acc1_Stress3_T0-5_A acc1_Ct_T0 acc1 Stress3 T0.5 A run1
acc1_Stress3_T0-5_B acc1_Ct_T0 acc1 Stress3 T0.5 B run1
acc1_Stress3_T0-5_C acc1_Ct_T0 acc1 Stress3 T0.5 C run1
acc1_Stress3_T1_A acc1_Ct_T0 acc1 Stress3 T1 A run1
acc1_Stress3_T1_B acc1_Ct_T0 acc1 Stress3 T1 B run1
acc1_Stress3_T1_C acc1_Ct_T0 acc1 Stress3 T1 C run1
acc1_Stress3_T2_A acc1_Ct_T0 acc1 Stress3 T2 A run1
acc1_Stress3_T2_B acc1_Ct_T0 acc1 Stress3 T2 B run1
acc1_Stress3_T2_C acc1_Ct_T0 acc1 Stress3 T2 C run1
acc1_Stress3_T3_A acc1_Ct_T0 acc1 Stress3 T3 A run1
acc1_Stress3_T3_B acc1_Ct_T0 acc1 Stress3 T3 B run1
acc1_Stress3_T3_C acc1_Ct_T0 acc1 Stress3 T3 C run1
acc1_Stress4_T0-3_A acc1_Ct_T0 acc1 Stress4 T0.3 A run1

acc2_Ct_T0_A None acc2 None(control) Ct A run2
acc2_Ct_T0_B None acc2 None(control) Ct B run2
acc2_Ct_T0_C None acc2 None(control) Ct C run2
acc2_Stress2_T0-2_A acc2_Ct_T0 acc2 Stress2 T0.2 A run2
acc2_Stress2_T0-2_B acc2_Ct_T0 acc2 Stress2 T0.2 B run2
acc2_Stress2_T0-2_C acc2_Ct_T0 acc2 Stress2 T0.2 C run2
acc2_Stress2_T0-3_A acc2_Ct_T0 acc2 Stress2 T0.3 A run2
acc2_Stress2_T0-3_B acc2_Ct_T0 acc2 Stress2 T0.3 B run2
acc2_Stress2_T0-3_C acc2_Ct_T0 acc2 Stress2 T0.3 C run2
acc2_Stress2_T0-7_A acc2_Ct_T0 acc2 Stress2 T0.7 A run2
acc2_Stress2_T0-7_B acc2_Ct_T0 acc2 Stress2 T0.7 B run2
acc2_Stress2_T0-7_C acc2_Ct_T0 acc2 Stress2 T0.7 C run2
acc2_Stress2_T1_A acc2_Ct_T0 acc2 Stress2 T1 A run2
acc2_Stress2_T1_B acc2_Ct_T0 acc2 Stress2 T1 B run2
acc2_Stress2_T1_C acc2_Ct_T0 acc2 Stress2 T1 C run2
acc2_Stress3_T0-2_A acc2_Ct_T0 acc2 Stress3 T0.2 A run2
acc2_Stress3_T0-2_B acc2_Ct_T0 acc2 Stress3 T0.2 B run2
acc2_Stress3_T0-2_C acc2_Ct_T0 acc2 Stress3 T0.2 C run2
acc2_Stress3_T0-3_A acc2_Ct_T0 acc2 Stress3 T0.3 A run2
acc2_Stress3_T0-3_B acc2_Ct_T0 acc2 Stress3 T0.3 B run2
acc2_Stress3_T0-3_C acc2_Ct_T0 acc2 Stress3 T0.3 C run2
acc2_Stress3_T0-7_A acc2_Ct_T0 acc2 Stress3 T0.7 A run2
acc2_Stress3_T0-7_B acc2_Ct_T0 acc2 Stress3 T0.7 B run2
acc2_Stress3_T0-7_C acc2_Ct_T0 acc2 Stress3 T0.7 C run2
acc2_Stress3_T1_A acc2_Ct_T0 acc2 Stress3 T1 A run2
acc2_Stress3_T1_B acc2_Ct_T0 acc2 Stress3 T1 B run2
acc2_Stress3_T1_C acc2_Ct_T0 acc2 Stress3 T1 C run2


ADD REPLYlink modified 5 months ago • written 5 months ago by uguy0

"The time points are not the same across experiments"

Because the time points weren't the same across experiments, and you have replicates for each time point, I'd suggest analyzing the two datasets separately. This makes sense for a number of reasons.

I only asked about the design, because if you had the same time points, and no replicates, then you would want to combine them, to provide some amount of replication.

ADD REPLYlink written 5 months ago by Michael Love19k

I understand your point.  But I then want to use a method that takes reads counts as input and I wonder if it's a problem that the sequencing depth is not normalized between the two datasets. (no comparison against the control reads counts).

ADD REPLYlink written 5 months ago by uguy0

Can you name a specific comparison you want to make that goes across the runs? You can’t directly compare across batches because you can’t separate the biological differences from technical. But it is possible to compare LFC across (comparing each to control first).

ADD REPLYlink written 5 months ago by Michael Love19k
gravatar for Michael Love
5 months ago by
Michael Love19k
United States
Michael Love19k wrote:

Can you show the experimental design and indicate what kind of biological replication there is? Do they have the same time points? Do you want to analyze them together, e.g. produce a results table which summarizes both into one?

ADD COMMENTlink written 5 months ago by Michael Love19k
Please log in to add an answer.


Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.2.0
Traffic: 347 users visited in the last hour