On 2014 Jan 12, Christian Frech commented:

GPHMM authors comment on their project home page (http://bioinformatics.ustc.edu.cn/gphmm/) on the performance of GPHMM in this comparison:

"We would like to point out critical errors found in a recently paper "Comparison of methods to detect copy number alterations in cancer using simulated and real genotyping data" published in BMC Bioinformatics with PMID 22870940, which was trying to compare different computational approaches including the GPHMM method.

As demonstrated in the GPHMM paper, we showed the superior performance of GPHMM on a cell-line dataset (see Figure 1,2,3 and Table 2 in the paper). However in this BMC paper, we found that the authors evaluated our method with the same dataset, but claimed that GPHMM failed to recognize the alteration pattern in the cell-line samples. Astonished by this apparent contradiction, we contacted them and later it became clear that during the test they WRONGLY replaced an important data file (“hhall.hg18_m.pfb”) in the tool package with another file used in their study.

Given the fact that this is a survey paper trying to accurately compare different methods and provide unbiased guidance to readers and the conclusion they made will pose influence on user's final choice of method in their studies, we suggested them to retest GPHMM and update their results. Unfortunately, except an ambiguous statement in the text saying “the baseline shift can be correctly estimated if a PFB with a modified specification is used”, we found the results and conclusion were not changed at all in the final version of this paper. Therefore, we argue that the performance of GPHMM is significantly underestimated in this paper."

On 2014 Jan 12, Christian Frech commented:

GPHMM authors comment on their project home page (http://bioinformatics.ustc.edu.cn/gphmm/) on the performance of GPHMM in this comparison:

"We would like to point out critical errors found in a recently paper "Comparison of methods to detect copy number alterations in cancer using simulated and real genotyping data" published in BMC Bioinformatics with PMID 22870940, which was trying to compare different computational approaches including the GPHMM method.

As demonstrated in the GPHMM paper, we showed the superior performance of GPHMM on a cell-line dataset (see Figure 1,2,3 and Table 2 in the paper). However in this BMC paper, we found that the authors evaluated our method with the same dataset, but claimed that GPHMM failed to recognize the alteration pattern in the cell-line samples. Astonished by this apparent contradiction, we contacted them and later it became clear that during the test they WRONGLY replaced an important data file (“hhall.hg18_m.pfb”) in the tool package with another file used in their study.

Given the fact that this is a survey paper trying to accurately compare different methods and provide unbiased guidance to readers and the conclusion they made will pose influence on user's final choice of method in their studies, we suggested them to retest GPHMM and update their results. Unfortunately, except an ambiguous statement in the text saying “the baseline shift can be correctly estimated if a PFB with a modified specification is used”, we found the results and conclusion were not changed at all in the final version of this paper. Therefore, we argue that the performance of GPHMM is significantly underestimated in this paper."