tuningParameters1.Rmd

---
title: "Tuning parameters for L-shape selection methods"
output:
  html_document:
    highlight: textmate
    theme: spacelab
    toc: yes
  html_notebook: default
---

# Introduction

Selection of L-shaped genes is being approached based on different methods, which depend on a variety of parameters. Changing the parameters affects the number of genes taht will be called "L-shaped" so we would want to find an optimal set of parameters for every method such that some performance measures such as sensitivity or specificity can be optimized.

A problem with this approach is that it requires a set of "TRUE Positives" -genes known to have L-shape- or TRUE negatives genes kown not to have L-shape- and this is very hard to obtain.

A way to do it is by visual inspecting the genes and selecting two sets that can be described as "clearly _" and "clearly non-L" which, by the way, is a more difficult concept.


```{r message=FALSE}
options(digits=4)
workingDir <- getwd()
dadesDir <-  file.path(workingDir, "dades")
codeDir <-  file.path(workingDir, "Rcode")
resultsDir <- file.path(workingDir, "results")
```


```{r include= FALSE}
installBiocifnot <- function(pckgName){
  if (!(require(pckgName, character.only = TRUE))) {
    source("http://Bioconductor.org/biocLite.R")
    biocLite(pckgName)
    require(pckgName, character.only = TRUE)
  }
}
# Example 
# installBiocifnot("CMA")

installifnot <- function(pckgName){
  if (!(require(pckgName, character.only = TRUE))) {
    install.packages(pckgName, dep = TRUE)
    require(pckgName, character.only = TRUE)
  }
}
# Example 
# installifnot("xlsx")

installGitifnot <- function(pathGit, pckgName, proxy = FALSE, urlproxy = "conf_www.ir.vhebron.net", portproxy = 8081, force.install = FALSE){
  if (!(require(pckgName, character.only = TRUE)) | force.install) {
    installifnot("devtools")
    # require(curl)
    if (proxy) set_config(use_proxy(url = urlproxy, port = portproxy)) 
    install_github(file.path(pathGit,pckgName), force = force.install)
    require(pckgName, character.only = TRUE)
  }
}
# Example 
# installGitifnot("miriammota","mmotaF",force.install = TRUE)
```

```{r}
installifnot("WriteXLS")
```

```{r message= FALSE}
source("Rcode/correlationFunctions.R")
source("Rcode/cmiFunctions.R")
source("Rcode/gridFunctions.R")
```

# TRUE and FALSE L-shaped genes

We have selected 90 genes calling them arbitrarily "TRUE" or "FALSE" L-shaped based on visual inspection.

```{r, eval=FALSE}
# Esta parte se ha hecho una vez
require(readxl)
DArangoTrueFalse <- read_excel("dades/DArangoTrueFalse.xls")
DArangoTrueFalse <- as.data.frame(DArangoTrueFalse)
TRUELGenes<- !is.na(DArangoTrueFalse$Mostreig_Sistematic) & DArangoTrueFalse$Mostreig_Sistematic=="TRUE"
trueLGeneNames <- DArangoTrueFalse[TRUELGenes,"Gen"]
FALSELGenes <- !is.na(DArangoTrueFalse$Mostreig_Sistematic) & DArangoTrueFalse$Mostreig_Sistematic=="FALSE"
falseLGeneNames <- DArangoTrueFalse[FALSELGenes,"Gen"]
falseLGeneNames <- c(falseLGeneNames, "ADA", "ANXA13", "BMPER", "DNAJA4", "ELF1", "GRAMD3", "LDHB" )
write.table(trueLGeneNames, file="dades/genesTrueLNEW.txt", row.names = FALSE, quote=FALSE, col.names = FALSE)
write.table(falseLGeneNames, file="dades/genesFalseLNEW.txt", row.names = FALSE, quote=FALSE, col.names = FALSE)
```

After some preprocessing genes have been written to two text files from where they are recovered.

```{r}
trueLGeneDF <-read.table("dades/genesTrueLNEW.txt")
(trueLGeneNames <- as.character(trueLGeneDF[,1]))
falseLGeneDF <- read.table("dades/genesFalseLNEW.txt")
(falseLGeneNames <- as.character(falseLGeneDF[,1]))
```


The list of FALSE and TRUE L Genes can be confronted with their scores to see up to what point the systems work decently well.

```{r}
DAscores <- read.csv( file="results/LshapeScoresDA.csv")
rownames(DAscores) <- DAscores$X
DAscores<-DAscores[,-1]

trueLScores <- DAscores[rownames(DAscores) %in% trueLGeneNames ,]
falseLScores <- DAscores[rownames(DAscores) %in% falseLGeneNames ,]
require(WriteXLS)
WriteXLS(c("DAscores", "trueLScores", "falseLScores"), ExcelFileName = "results/DAScoresNEW.xls", row.names = TRUE)
```