This is a test version of Biostars. For the public version, visit https://www.biostars.org.
How to I perform MDS clustering on microarray data to identify gender discrepancies between samples?

Hi all, I recently joined a lab and I am currently learning how to perform quality control on microarray gene expression data. I am very new at this and I have been following the methodology used by a previous project at the lab (according to Item 3.2 here).

I am currently trying to perform MDS clustering on my microarray gene expression data in order to separate samples based on their gender. Based on the methods, I am supposed to be able to see two distinct clusters. However, when trying it out on my own in R using limma, I can only acquire a single cluster. How exactly should I go about doing this?

Here is my R code:

library(limma)
expression_data <- t(rma_transposed) # my gene expression data after RMA, with samples in columns and probesets in rows
mds <- plotMDS(expression_data, gene.selection = "common")
plot(mds)

The resulting plot has only a big cluster in the middle, whereas the 'correct' plot has two distinct clusters. Does anyone have any suggestions on what I'm doing wrong? Thank you!

microarray gene expression gender sample
• 242 views • link
written by Volka Volka

Did you perform MDS on the "top 500" probes? Here is the text from the manuscript:

Sample differentiation was examined and was achieved through the use of clustering using the program “plotMDS” from the limma package based on the top 500 transcript probes (Supplementary Method 4.3). We observed two distinct gender clusters from the MDS plot. (iii) samples with gender discordance between recorded genders and MDS plot (2 samples were removed).

• 1 views • link
written by Kevin Blighe Kevin Blighe

Hi thanks for the reply. Would that mean using the "top=500" argument for limma's plotMDS? Or does that mean something else entirely, like manually extracting the top 500 probes?

• 1 views • link
written by Volka Volka

0 answers

No answers yet.

Log in to answer this question.

More posts like this