Reader Comments
Post a new comment on this article
Post Your Discussion Comment
Please follow our guidelines for comments and review our competing interests policy. Comments that do not conform to our guidelines will be promptly removed and the user account disabled. The following must be avoided:
- Remarks that could be interpreted as allegations of misconduct
- Unsupported assertions or statements
- Inflammatory or insulting language
Thank You!
Thank you for taking the time to flag this posting; we review flagged postings on a regular basis.
closeUse of “Silhouette” for Correlation Analysis
Posted by Charles_Warden on 06 Sep 2018 at 16:19 GMT
It has come to my attention that the use of the term “Silhouette” is not precise, in the context of the previous study by Verhaak et al. [1]. This is most noticeable in Figure 2, but the term “Silhouette” is used multiple times in the paper.
More specifically, we visualize significant correlation coefficients ordered by subtype and then by the value of the correlation coefficient. This is what we describe in the Methods, which includes a section called “Correlations with Signature Gene Expression (Silhouette plots)."
However, the Silhouette Plot in the Verhaak et al. paper refers to a distance metric for clusters. While we characterize correlations with the subtype genes (and enrichment of the genes for different types), these correlations are calculated separately for different datasets (so, each gene has more than one correlation) and we don’t assign subtypes based upon NMF clustering.
We provide interesting insights in how these signature genes relate to our candidate of interest (and patient characteristics), but we want to apologize for the confusion (since readers that only see Figure 2C/2D may initially think the plot represents the same information as the Verhaak et al. paper, and that interpretation would not be accurate).
This has been a very useful experience for me, and I hope that this comment can help others be mindful of reviewing terminology (and allocating enough time to carefully review papers being cited) when writing their research articles.
[1] - Verhaak et al. (2010) “Integrated genomic analysis identifies clinically relevant subtypes of glioblastoma characterized by abnormalities in PDGFRA, IDH1, EGFR, and NF1.” Cancer Cell: 17(1):98-110
RE: Use of “Silhouette” for Correlation Analysis
Charles_Warden replied to Charles_Warden on 07 Sep 2018 at 18:34 GMT
I am very sorry, but I later realized that I mixed up the methods between some papers (the Verhaak et al. paper doesn’t use NMF clustering).
So, the section of the comment above saying:
“However, the Silhouette Plot in the Verhaak et al. paper refers to a distance metric for clusters…and we don’t assign subtypes based upon NMF clustering.”
Would be accurately described as:
“However, the Silhouette Plot in the Verhaak et al. paper refers to a distance metric for clusters…and we don’t assign subtypes to any new samples in this paper.”
I apologize again for the confusion.