Interpreting the proximity matrix

[Paul o Reilly]

Hi,

I constructed a randomForest model with 2 different classes (drug sensitive and drug resistant cell lines) using protein expression values. This model had a low OOB error of ~5% and when used to predict drug sensitivity on an independent test set it performed well (Area under the receiver operator curve=0.85). We are now trying to show that the model is describing multiple subtypes of resistant cell lines. 

We are beginning this process by looking at the proximity matrix. My question would be can we perform clustering on this proximity matrix to identify those groups of cell lines close to each other and if so does anyone have any suggestions what approach would be best ?

Yours

Paul