Question

DESEQ2 normalize data of population data

0

Entering edit mode

vrehaman • 0

@vrehaman-14957

Last seen 4.0 years ago

Dear Team,

I want to perform differential expression on counts generated from featureCounts. Each family has different disease. Can we perfrom differential expression across all samples? or since disease is different from family to family, should we perform separately family by family?

Example sample information is like below

SampleID	condition	Family
Sample1	Normal	Fam1
Sample2	Diseased	Fam1
Sample3	Normal	Fam2
Sample4	Diseased	Fam2
Sample5	Normal	Fam3
Sample6	Diseased	Fam3
Sample7	Normal	Fam4
Sample8	Diseased	Fam4
Sample9	Normal	Fam5
Sample10	Diseased	Fam5

I followed the below link for grouping condition and family and get multiple comparisons. Is it ok t proceed like this?

DESEq2 comparison with mulitple cell types under 2 conditions

dds <- DESeqDataSetFromMatrix(countData = counts, colData = coldata , design = ~ Family + condition)

dds$group <- factor(paste0(dds$Family, dds$condition))

design(dds) <- ~ group
dds <- DESeq(dds)
resultsNames(dds)

Could you please let me know your suggestions.

Thanks In Advance

Fazulur Rehaman

deseq2 multiple factor design • 1.2k views

ADD COMMENT • link updated 5.6 years ago by Michael Love 41k • written 5.6 years ago by vrehaman • 0

score 1 · Answer 1 · 2018-10-08

1

Entering edit mode

Michael Love 41k

@mikelove

Last seen 1 day ago

United States

What kind of genes are you looking for? Genes commonly DE in diseased relative to normal?

You cannot perform DE within family here, because there are no "replicates" (I don't know if these are human donors, or what exactly a sample refers to...).

ADD COMMENT • link 5.6 years ago Michael Love 41k

0

Entering edit mode

Dear Michael,

Thanks a lot for your quick response.

Please find below the details

What kind of genes are you looking for? Genes commonly DE in diseased relative to normal?

We are looking for more diabetic and obesity genes commonly DE in diseased relative to normal.

You cannot perform DE within family here, because there are no "replicates" (I don't know if these are human donors, or what exactly a sample refers to...).

yes, these are human donors having diabetic or obseity. Each family has one control and one diseases sample (more than 6 families). Disease might be either diabetic or obesity.

Please suggest me how can I proceed with DE.

Thanks In Advance

Fazulur Rehaman

ADD REPLY • link 5.6 years ago vrehaman • 0

1

Entering edit mode

You can use ~family + condition, which will control for family baseline, while finding genes where the diseased samples show DE relative to normal.

ADD REPLY • link 5.6 years ago Michael Love 41k

0

Entering edit mode

Dear Michael,

Thanks a lot for your suggestions.

At first, I used the same model ~family + condition

Here are the details:

which will control for family baseline, while finding genes where the diseased samples show DE relative to normal.

It means only one comparison where we can find genes diseased vs Normal, irrespective of which disease, right?

> resultsNames(dds)

[1] "Intercept" "family_Fam139_vs_Fam10"

[3] "family_Fam193_vs_Fam10" "family_Fam43_vs_Fam10"

[5] "family_Fam52_vs_Fam10" "family_Fam53_vs_Fam10"

[7] "family_Fam55_vs_Fam10" "family_Fam8_vs_Fam10"

[9] "condition_Normal_vs_Diseased"

Building Results table for diseased vs Normal condition.

res1 <- results(dds, contrast=c("condition","Diseased","Normal"))

In resultsNames, it was mentioned as "condition_Normal_vs_Diseased". Since we have to get genes DE in diseased relative to normal, in results() function I have given "Diseased" followed by the "Normal". Please confirm if it is ok?

And also resultsNames() function giving other possible contrasts which are "family_Fam139_vs_Fam10", `"family_Fam43_vs_Fam10" etc. How can I use them.

Please let me know your suggestions.

Thanks In Advance

Fazulur Rehaman

ADD REPLY • link 5.6 years ago vrehaman • 0

1

Entering edit mode

Take a look at the DESeq2 vignette note on factor levels, where this is discussed.

But in short, what you have above is fine, you can always specify the LFC you want using contrast. The object res1 is your results table of interest.

You do not use the other coefficients listed in resultsNames, they are nuisance coefficients for your purposes. Those coefficients control for the family baselines.

ADD REPLY • link 5.6 years ago Michael Love 41k

0

Entering edit mode

Dear Michael,

Thanks a lot for your confirmation and suggestions on factor levels.

Thanks & Regards

Fazulur Rehaman

ADD REPLY • link 5.6 years ago vrehaman • 0

0

Entering edit mode

Dear Michael,

I have one more question. Since, we have only one comparsion which is diseased relative to Normal, Is there any possibility, I might know the family belongs to upregulated or down regulated genes.

Please let me know.

Thanks & Regards

Fazulur Rehaman

ADD REPLY • link 5.6 years ago vrehaman • 0

0

Entering edit mode

You can look at plotCounts()

ADD REPLY • link 5.6 years ago Michael Love 41k