setwd("/Users/lszabo/Documents/RProjects/ButteLab/biomarkers")
source("forestplot.R")
## Loading required package: rmeta
## Loading required package: grid
load("Ovarian_Cancer_Analysis.RData") #all data objects from meta-analysis
Genes in Mor et al. 2005 proposed early biomarkers
This is an example they picked based on initial analysis but it did not validate. Our meta-analysis results are consistent with this since it was detected in some datasets but not others.
forest.plot(allData$output.REM, key = "EGF")
Now look at the 4 genes they included in their final recommendation.
gene 1: OPN, this one did show up in meta-analysis but not significant between normal/early in TCGA
forest.plot(allData$output.REM, key = "SPP1")
The rest of their 4 markers did not show up in our meta-analysis
gene 2: Prolactin - this is up in all of the studies, but the effect size was not very large. We filtered out anything with a log2 value of less than 0.75 because we had plenty with bigger effect sizes.
forest.plot(allData$output.REM, key = "PRL")
gene 3: Leptin - barely up in most of our studies, but not in 2.
forest.plot(allData$output.REM, key = "LEP")
gene 4: IGF-II - up in most studies but significantly down in 1. So could argue that this is a reasonable candidate and something is wrong in that 1 dataset, but we do have plenty of genes with a larger effect size in all 7 studies.
forest.plot(allData$output.REM, key = "IGF2")
In table 1, only CA-125, MIF, and OPN had p < .000001. Of these, CA-125 and OPN are in our list of top 160 genes from meta-analysis.
The mean patient value for MIF was up in all datasets, but a large number of samples had decreased MIF so did not meet our FDR criteria.
forest.plot(allData$output.REM, key = "GIF")
Visintin et al. 2008 added 2 more genes to list from Mor et al. (Ovasure product)
gene 1: MIF/GIF is shown and discussed above.
gene 2: CA-125/MUC16 is picked up as one of the top genes in our meta-analysis (#13 on our list of 160). But it is not significantly different between normals and early stage in TCGA cohort. This is not surprising since CA-125 is known to be a poor marker for early stage.
forest.plot(allData$output.REM, key = "MUC16")
