A non-BioC question: Does anyone know of an authoritative implementation of the van 't Veer et al (2002) breast cancer metastasis classifier? There seem to be different interpretations in the literature, in some places (e.g., van de Vijver 2002, Tibshirani & Efron 2002, Mansmann et al2006) it's described as classifying each sample according to whether the correlation with the centroid of good prognosis cases is above some threshold, and in others the classification is the class with the centroid that has highest correlation with the sample (which is how I interpret van 't Veer 2002). Thanks, Gad -- Gad Abraham MEng Student, Dept. CSSE and NICTA The University of Melbourne Parkville 3010, Victoria, Australia email: gabraham at csse.unimelb.edu.au web: http://www.csse.unimelb.edu.au/~gabraham

Hi Gad, I don't know of a reference implementation for this. The differences you explain depend on whether you want a more specific or a more sensitive classifier. As you mentioned, different variations have been used in the different papers about it. Just decide what you want to do (maximizing accuracy is generally a decent choice) or choose one publications and you'll be fine. Francois Gad Abraham wrote: > A non-BioC question: > > Does anyone know of an authoritative implementation of the van 't Veer > et al (2002) breast cancer metastasis classifier? > > There seem to be different interpretations in the literature, in some > places (e.g., van de Vijver 2002, Tibshirani & Efron 2002, Mansmann et > al2006) it's described as classifying each sample according to whether > the correlation with the centroid of good prognosis cases is above some > threshold, and in others the classification is the class with the > centroid that has highest correlation with the sample (which is how I > interpret van 't Veer 2002). > > Thanks, > Gad >