On 2015 Dec 03, Robert M Flight commented:

I have some concerns about the procedure of propagation to GO child terms mentioned as part of the procedure for generating the mutation-GO profile, as this results in new gene-GO annotations that do not exist in the original gene-GO annotation. A github repo with the results of my investigation of this issue is available, along with copies of the paper, supplementary data, all R and octave code used, and my correspondence with Sael.

The authors state that this propagation is necessary for the function of the ONMF procedure, but do not provide any evidence that this is so, and as part of the overall analysis actually perform propagation back to parent terms to determine significant GO terms. I wonder if both sets of propagation might have been avoided by using a GOSlim or another GO subset.

On 2015 Dec 03, Robert M Flight commented:

I have some concerns about the procedure of propagation to GO child terms mentioned as part of the procedure for generating the mutation-GO profile, as this results in new gene-GO annotations that do not exist in the original gene-GO annotation. A github repo with the results of my investigation of this issue is available, along with copies of the paper, supplementary data, all R and octave code used, and my correspondence with Sael.

The authors state that this propagation is necessary for the function of the ONMF procedure, but do not provide any evidence that this is so, and as part of the overall analysis actually perform propagation back to parent terms to determine significant GO terms. I wonder if both sets of propagation might have been avoided by using a GOSlim or another GO subset.