filtering of genes before unsupervized analysis

Question

[kanwarjag]

I have an expression array data and want to from PCA or heat map. Do we filter noninformative genes such as no expression or zero counts in RNAseq before doing any of such analysis? Why it is necessary to filter such genes.

Thanks

Answer 1

[ATpoint]

Arrays measure relative differences between expression, not absolute values. Therefore you cannot infer if a gene is highly-or lowly expressed. Arrays only inform you about differences between conditions. For information about expression levels per sample you need RNA-seq. For heatmaps you could use the log2-fold changes between conditions of the DEG genes of your arrays. For PCA see What values should I use for principle component analysis of microarray data?