Get list of gene signatures published

Question

[sabin]

Hello everyone, I need to get the list of gene signatures defined from Bailey et al. and used for tumor subtyping, published in the study:

Bailey P, Chang DK, Nones K, et al. Genomic analyses identify molecular subtypes of pancreatic cancer. Nature. 2016;531(7592):47-52. doi:10.1038/nature16965

However, from the publication I couldn't find neither such list of genes nor how many these genes are. In other studies, where this list of genes is used as reference or for comparison, the authors say that the genes are 707. But still, I didn't find the list of these 707 genes anywhere. If the genes have been used in other studies then they must be somewhere, however i can't find then anywhere.

Anyone knows how to get them?

Thanks in advance!

[ATpoint]

Hi, did you check the supplementary materials for tables that contain these genes, or see whether they reference a paper that lists them? If neither is successful then you will need to email the authors and hope for a response. That's the choices you have.

[sabin]

Hi! thanks for the reply, Yes, I've checked all the tables and supplementary material that i could find, but nothing. I'll see if anyone on the forum worked with these genes, in the meantime i text the authors too. Best, Manuela

[igor]

The genes are in the Supplementary Table 14, which is actually a little confusing because it's an Excel spreadsheet with multiple sheets, so it's not really a single table.

[sabin]

Hi Igor, Yeah i agree, I guess those are the genes. I was confused because in other papers i found different numbers for what concern the amount of genes, so i wasn't sure which was the correct one. Also, in the suppl Table 14 there are genes obtained from multiclass analysis and then other sheets that, according to their description in the paper, they are the genes classifier for each subtypes. So that is also confusing me about what to use to classify the tumor samples, if the genes got from the multiclass analysis (differentially expressed genes across their defined classes) or if i have to merge the genes classifier of each subtype obtained with an analysis of OneSubtypeVStheRest. Thank you for the help!

[igor]

Which papers give a different number of genes?

From the paper:

Gene sets representing PDAC classes were generated by selecting significantly upregulated genes in a given class versus all other classes. An adjusted P value of 0.01 was used as the cut-off in each case.

That seems to be referring to "OneSubtypeVStheRest".

[sabin]

Hi Igor, I found these 2 papers (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5964983/, https://www.biorxiv.org/content/10.1101/690552v1.full.pdf) where they use 613 genes (the ones obtained from multiclass analysis, so from the excel sheet "PDAC_sam_delta_9_8"). In this paper: https://molecular-cancer.biomedcentral.com/articles/10.1186/s12943-017-0739-z#Sec3, they talk abut 707 genes. In this other paper: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5606023/, they talk about 859 genes.

That's what confuses me about the right list of genes to consider. Best, Manuela

[igor]

Thanks for sharing these. You're right. They are all different. I would trust the TCGA paper the most since many people were involved. They specifically mention that 463 of 613 genes overlapped with their dataset, which would usually be my first guess why the numbers differ between different sources. Based on the data from the original paper, it seems like it should be almost twice as many.

[sabin]

Thanks a lot for your help, Igor! I would also go with the 613. Best, Manuela

[igor]

Please post a follow-up if you figure out exactly how to get to 613.

[sabin]

Hi Igor,

I got the 613 genes from suppl table 14, in the excel sheet "PDAC_sam_delta_9_8".

[igor]

You're right. That number of genes matches exactly. Were you able to figure out what that table actually represents?

[sabin]

As far as i understood from the paper, those genes should be genes differentially expressed between all the classes.

[igor]

That sounds reasonable. but what do sam, 9, and 8 refer to?

[sabin]

In the paper they say

To define genes differentially expressed between all classes we used the function ‘sam’ as implemented in the R package ‘siggenes’.

Where SAM stands for Significance Analysis of Microarrays. About the "8" and "9" I couldn't find anything :/

[igor]

It's unfortunate that the main output of the paper is not more clearly described.

Curiously, not all of those 613 genes are present in the "OneSubtypeVStheRest" tables.