0
$\begingroup$

This question was also asked on BioStars

I am trying to recreate this heatmap. How would I compare the four variables: Ly49+/- and MOG/MOGSP to get a single heatmap? Thank you for helping me through this.

I need to compare both Ly49+ versus Ly49− and MOG versus MOG plus SP. Column data in at the bottom. The function I used for DESeq2 is,

dds = DESeqDataSetFromMatrix(countData=countData,

colData=colData,

design= ~ cell_treatment)

I am not sure if the design is right. For the results, how should I write the contrasts? I am very confused and would really appreciate any help!

DataFrame with 11 rows and 2 columns

cell treatment

<character> <character>

MOGSP_2 SPtetramer+CD8+TCell MOGSP

MOGSP_3 SPtetramer+CD8+TCell MOGSP

MOG_3 SPtetramer+CD8+TCell MOG

MOG_4 SPtetramer+CD8+TCell MOG

MOG_5 SPtetramer+CD8+TCell MOG

Ly49N_1 Ly49-CD8+TCell MOGSP

Ly49P_1 Ly49+CD8+TCell MOGSP

Ly49N_2 Ly49-CD8+TCell MOGSP

Ly49P_2 Ly49+CD8+TCell MOGSP

Ly49N_3 Ly49-CD8+TCell MOGSP

Ly49P_3 Ly49+CD8+TCell MOGSP

Thank you!!

$\endgroup$
2
  • 2
    $\begingroup$ Cross-posted: biostars.org/p/461091 $\endgroup$
    – ATpoint
    Sep 13 '20 at 6:35
  • 1
    $\begingroup$ I closed the cross-post at Biostars because @StupidWolf basically asked for the same details as I did there, and there is no need to split information across two communities. $\endgroup$
    – ATpoint
    Sep 14 '20 at 16:24
3
$\begingroup$

Your table is not very clear. If I understood it correctly, there are three groups of t-cells, SPtetramer+, Ly49- and Ly49+ and within it there is a MOGSP vs MOG comparison. I would set colData like this:

                   name             Ly treatment
1  SPtetramer+CD8+TCell SPtetramerplus     MOGSP
2  SPtetramer+CD8+TCell SPtetramerplus     MOGSP
3  SPtetramer+CD8+TCell SPtetramerplus       MOG
4  SPtetramer+CD8+TCell SPtetramerplus       MOG
5  SPtetramer+CD8+TCell SPtetramerplus       MOG
6        Ly49-CD8+TCell      Ly49minus     MOGSP
7        Ly49+CD8+TCell       Ly49plus     MOGSP
8        Ly49-CD8+TCell      Ly49minus     MOGSP
9        Ly49+CD8+TCell       Ly49plus     MOGSP
10       Ly49-CD8+TCell      Ly49minus     MOGSP
11       Ly49+CD8+TCell       Ly49plus     MOGSP

I generate an example dataset:

counts = matrix(rnbinom(11000,mu=100,size=1),ncol=11)

And in the design you specify these two factors, meaning gene expression can be affected by these two conditions:

dds = DESeqDataSetFromMatrix(counts,colData=colData,~Ly +treatment)
dds = DESeq(dds)

You can compare them like this:

head(results(dds,c("Ly","Ly49plus","Ly49minus")))
log2 fold change (MLE): Ly Ly49plus vs Ly49minus 
Wald test p-value: Ly Ly49plus vs Ly49minus 
DataFrame with 6 rows and 6 columns
          baseMean     log2FoldChange            lfcSE               stat
         <numeric>          <numeric>        <numeric>          <numeric>
1 91.4458114779915  0.712987165797688 1.28291188260395  0.555756927241586
2 81.7945234843954 -0.418266530066077 1.14378290088875 -0.365686993345568
3 94.4351681904941 -0.775592496878288 1.32965355129901 -0.583304196886903
4 107.484335203298  -1.36694885973712 1.38595911743023 -0.986283680770933
5 61.2205696982841 -0.760219864564472 1.16779888688769 -0.650985262188887
6 95.5908168437137  0.497427543096335 1.12768919469662   0.44110340458672
             pvalue              padj
          <numeric>         <numeric>
1 0.578377034355333 0.977537706056193
2    0.714598652706 0.992321035598301
3 0.559688537757548 0.977537706056193
4 0.323993925580938 0.954772990635094
5 0.515056000552832 0.977537706056193
6 0.659138138867294 0.992321035598301

To compare groups within another factor:

results(dds,c("treatment","MOGSP","MOG"))
log2 fold change (MLE): treatment MOGSP vs MOG 
Wald test p-value: treatment MOGSP vs MOG 
DataFrame with 6 rows and 6 columns
          baseMean     log2FoldChange            lfcSE               stat
         <numeric>          <numeric>        <numeric>          <numeric>
1 91.4458114779915 -0.875905107403943 1.43007567177694 -0.612488642867122
2 81.7945234843954  0.595588650280833 1.28109080878515  0.464907441530728
3 94.4351681904941 0.0255578400583986 1.48592401497315 0.0171999643325371
4 107.484335203298  -1.46652057431159 1.55285829334496 -0.944400774105796
5 61.2205696982841  -1.27442596494316 1.30489822606005  -0.97664778715433
6 95.5908168437137  -2.35402796630456 1.26608624796739  -1.85929510733078
              pvalue              padj
           <numeric>         <numeric>
1  0.540214509839105 0.944881849859735
2  0.641997741235223 0.973651054534338
3  0.986277090644156 0.994944323499595
4  0.344964886056295  0.87421063696937
5  0.328743552427222                NA
6 0.0629853201285643 0.525889480109579

The colData i used:

structure(list(name = structure(c(3L, 3L, 3L, 3L, 3L, 1L, 2L, 
1L, 2L, 1L, 2L), .Label = c("Ly49-CD8+TCell", "Ly49+CD8+TCell", 
"SPtetramer+CD8+TCell"), class = "factor"), Ly = c("SPtetramerplus", 
"SPtetramerplus", "SPtetramerplus", "SPtetramerplus", "SPtetramerplus", 
"Ly49minus", "Ly49plus", "Ly49minus", "Ly49plus", "Ly49minus", 
"Ly49plus"), treatment = structure(c(2L, 2L, 1L, 1L, 1L, 2L, 
2L, 2L, 2L, 2L, 2L), .Label = c("MOG", "MOGSP"), class = "factor")), class = "data.frame", row.names = c(NA, 
-11L))
$\endgroup$
5
  • $\begingroup$ so it can used for multiple variable or cell type comparisons instead of many pairwise comparison which i do normally $\endgroup$
    – kcm
    Sep 16 '20 at 5:27
  • 1
    $\begingroup$ it depends on the design but in general yes. In this example, you can see it's necessary because the groups have some overlap $\endgroup$
    – StupidWolf
    Sep 16 '20 at 16:15
  • $\begingroup$ conceptual doubt so if i have to perform multiple pairwise comparison do i need to set reference level for every comparison or i can simply use "results(dds,c("treatment","MOGSP","MOG"))" or the contrast function? $\endgroup$
    – kcm
    Jan 21 at 6:36
  • 1
    $\begingroup$ using results(dds,c("treatment","MOGSP","MOG")) takes care of everything. The reference level of the factor does not matter here $\endgroup$
    – StupidWolf
    Jan 22 at 1:09
  • $\begingroup$ will try this.. $\endgroup$
    – kcm
    Jan 22 at 11:12

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.