How to remove some known samples manually after hierarchical clustering observation ?

Question

[XBria]

Hi,

A basic question :) I have some samples to obtain DEG. After Hierarchical clustering, decided to remove a couple of samples. How may I (manually) remove them ? 1- before which process should I add the codes (before outlier detection? , before normalization ?) ? 2- Any tips or suggestions before adding the codes (in R) ?

Thanks

[ATpoint]

Can you give some context on why you want to do that? That probably helps advising strategies. Without context I guess you have some samples looking odd and now you are unsure if this is a technical issue or true biological variation? Best start with some quality control by PCA unless this has already been done.

Also what format do you start from, bam files, count matrix, quantification files? Please add some details.

Answer 1

[Jean-Karim Heriche]

Without details, we can't give specific answers. Removing samples will affect the downstream steps so consider why you want to remove them and based on this decide up to which point you can/want to keep them. If it's because of an experimental/technical issue then the samples should probably be discarded from the start, i.e. not even be considered for analysis. If it's because they are valid outliers then you may want to process them separately. The point at which you process them separately would depend on why they are outliers. They may show expression levels that could impact the normalization in which case you may want to separate them before the normalization step, Or they could represent a group with complex structure that you may want to cluster separately.