0
$\begingroup$

Brain data, particualrly human brain data have very heterogeneus cell types, so it's important to normalise by cell type/add proportoins of different cell types to the formula when performing differential expression. We have some bulk RNA-seq brain data from disease samples and wildtypes. I have used the package BrainInABlender to get a matrix of the proportions of each cell type in each sample

Cell type Sample_1 Sample 2
Neuron 0.5 0.7
Microglia 0.5 0.3

The rows are the types of cells, the columns are the samples. I wanted to know what's the best way to add this as a confounding variable in Enrichment Browser (or if that's not possible in DESeq2).

Edit: Current way I've tried to correct for this is in DESeq2. I have transposed the matrix of cell type proportoins then set DESeq's formula:

~ Condition + Condition:Neuron:Microglia:etc

This has drastically reduced the number of genes found to be differentially expressed but I don't know if it's the right way to do it.

Edit 2: When I tried removing the interaction terms, replacing them with +

~ Condition + Neuron + Microglia etc

As suggested on the Bioconductor support section the number of genes went up by an order of magnitude from just using condition which is the opposite of what should have happened. This persisted whether condition was at the start or end of my formula.

Improve this question
$\endgroup$
6
  • $\begingroup$ What have you tried? DESeq2 works with count data not decimal numbers. What is your biological question/goal? $\endgroup$ – llrs Jan 8 at 18:43
  • $\begingroup$ I have no idea what is an enrichment browser. If you are referring to RNAseq analysis, in DESeq2, when you set up the data with colData = ... just add the cell type neuron / microglia as a numeric column? $\endgroup$ – StupidWolf Jan 9 at 1:42
  • $\begingroup$ @llrs Sorry for not being clear eough. I have a seperate matrix of count data. I also have this matrix which has the proportoins of each cell type in each sample. My goal was to best correct for differences in proportoins of cell types in each sample. $\endgroup$ – Sethzard Jan 9 at 17:22
  • $\begingroup$ @StupidWolf Enrichment Browser is an R package that allows for running several differential expression methods easily. I have added it to the coldata and seperated with : in my formula (Condition:Neuron:Microglia:etc but have no idea if that's right. $\endgroup$ – Sethzard Jan 9 at 17:25
  • $\begingroup$ @Sethzard this doesn't solve the last question. Correct for differences in proportion of cell types for what? To see what cell type express? $\endgroup$ – llrs Jan 10 at 21:53

Your Answer

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

Browse other questions tagged or ask your own question.