Entering edit mode
9.1 years ago
Ron
★
1.2k
Hi all,
I am working on RNAseq expression analysis.I have 4 or 5 clusters of various samples.
I have this new sample and want to know that which cluster it is closely related to as seen in PCA.I have done hierarchical clustering as well but I would like to know a certain score if we see a "new sample" being closer to lets say 2 clusters.
Any suggestions for this?
Thanks in advance for assistance
-Ron
You mention both PCA and hierarchical clustering. PCA does not assign samples to clusters, so not clear what you mean by "clusters" with respect to PCA. Are you trying to assign your new sample to one of your previously defined clusters, or do you want to assign some score to each existing cluster, representing how "close" the new sample is to each of the clusters?
I want to assign some score to each existing cluster ,representing how close the new sample is to each of them.
For the hierarchical clusters then: I would compute the centroid of each cluster. Then, calculate the distance (using the same distance metric used in the clustering) from your new sample to each of the centroids. Those distances are then the scores, representing how close the new sample is to each of your clusters.