Entering edit mode
6.7 years ago
martin
▴
10
I recently started research in the field of (breast) cancer subtyping. My group is among other things interested in the molecular subtyping of triple negative breast cancer according to Lehman et al. (2011)
The group behind this paper also released a website tool (http://cbc.mc.vanderbilt.edu/tnbc/prediction.php) for analysis of own data sets. On the website there is an example data set provided and in the paper there are example results from public microarrays (Figure 4).
But I have a number of problems:
- I uploaded the example data set on the website and got an error (ER-positive samples).
- I removed the samples in question and the results do not match what is given on the help-pages of the website nor what is depicted in the second paper.
- I downloaded and normalized the publicly available microarray data sets used in the paper and upload them, in 8 of 25 cases the subtyping does not match with the paper.
- I plotted the expression values from the publicly available microarray data sets against the values given with the example file for the exact same samples to see if a data transformation was applied and the results show no correlation at all. (see below)
I would very thankful for any hint or - even better - code that could fill any of the gaps mentioned above.
Almost certainly a better option would be to contact the authors of the published work and/or the developers of the web tool.
I already contacted the dev of the webtool. I unfortunately got a non-answer.
Okay. I have also just tried the example data and it said:
Did you try to remove that sample to see if it even works?
Also, your gene symbols are in HUGO, right?
From original post:
Thanks genomax - was doing too many things at once here. I also sometimes add comments to threads that are likely to go unanswered in the hope that someone else sees the comment and jumps in.
martin, I did not want to say but these questions are likely only best answered by the TNBCtype developers, who appear to be unresponsive for whatever reason.
Here is my best take:
The data that they have made available is just an example; therefore, they may have applied an unknown level of filtering and transformations to it. As it's just sample data, I personally do not believe that they have to go through everything. It's purely for the purposes of showing you the correct data format and structure.
This is definitely for the authors involved. As it's published data, the methods should be very clear. The authors additionally have a duty to follow up on questions relating to their published work. If, however, funding for the project has ended and/or the people do not even work there anymore, then your options are limited.
Same as my first answer related to the sample data. Also make the consideration that functions for normalisation may have been modified since the work was first published, e.g., RMA versus GC-RMA
----------------------------------------
Finally, if you still get stuck, you could easily each out to one f the groups that have recently made a publication where they used TNBCtype:
https://scholar.google.com.br/scholar?client=ubuntu&um=1&ie=UTF-8&lr&cites=5956288328835640727
Thank you very much Kevin for your reply. I will definitely try to contact the original authors. Your probably right that the author of the web tool has maybe moved on. Since there are publications out there on TNBCtype, I was hoping that anyone in the community would find my request here. I will definitely follow up here with my findings, so the knowledge gets preserved. Thanks also to genomax for helping out with a short answer.
TNBC Subtyping tool is mainly for microarray data. Is your data RNA-seq? Did you get any reply from authors?