On 2013 Jun 15, Wolfgang Huber commented:

We performed a re-analysis of the same data using a variety of other, more standard classification procedures and could not reproduce the reported prediction accuracy. It is possible that the reported prediction accuracy was measured in a way that is sensitive to over-fitting, due to the high-dimensionality of the data and the choices that can be made for data "pre-processing". Our analysis was summarised in Ruschhaupt M, 2004 (pdf: http://www-huber.embl.de/pub/pdf/sagmb_2004.pdf).

On 2013 Jun 15, Wolfgang Huber commented:

We performed a re-analysis of the same data using a variety of other, more standard classification procedures and could not reproduce the reported prediction accuracy. It is possible that the reported prediction accuracy was measured in a way that is sensitive to over-fitting, due to the high-dimensionality of the data and the choices that can be made for data "pre-processing". Our analysis was summarised in Ruschhaupt M, 2004 (pdf: http://www-huber.embl.de/pub/pdf/sagmb_2004.pdf).