Selecting Samples based on Clustering

Question

[Hyper_Odin]

If I only select samples based on their clustering in heatmaps, am I not creating a bias ? For example, I had 35 samples (RNA Seq) nitially, but, in the end, we only selected 12 samples based on their clustering. When I checked for some of the genes in both cases, they were hugely varying in terms of P-adjusted value.

I mean, I know this is a bias but how do I explain this to my boss? Please Helpppp!!!

[rpolicastro]

Can you give us some information on what the samples are, your method for clustering, and how clusters were selected?

[Hyper_Odin]

So, there were 3 types of samples.

1. Treated Nodules - 16 samples
2. Control Nodules - 3 samples
3. Untreated Nodules - 16 samples

Performed Correlation clustering after DESEQ2 using Pheatmap

Now, let's say, only violet samples with large cluster was selected, and the cyan cluster was completely removed from the final analysis.