I am analyzing breast cancer samples to perform their subtype classification. For this purpose I have tested the genefu package, which allows for classification e.g. using PAM50 or AIMS. For the later, there is actually a separate package in R (called AIMS). The confusing thing is that results of the two packages differ, and they differ not only in terms of the assigned per-sample labels, but also in the prediction scores for each label. The genefu scores are much smaller, and the ones assigned with the AIMS package are rather unambiguous. Why is that? I have also run PAM50 with genefu, and the prediction scores are also low. Note: I have used exactly the same expression matrices for both packages, namely values that have been normalized and transformed for between sample comparison.
