######################################################################################
# Goal: mtCN vs TMT expression stepwise regression
#
# Input file: ccle176.mtCN_and_TMT.raw_vals.complete_obs.txt
#  columns: 176 cancer encyclopedia cell lines
#  rows: first row is mtCN (estimated from whole genome sequence data), other 5153 rows are proteins
#  values: TMT protein abundance metric of protein in cell line (or mtCN for first row)
#
# Contents:stepwise regression was performed using the R package leaps (regsubsets, nvmax=4, method="forward").
######################################################################################

library(tidyverse)
library(leaps)

# Input data with mtCN and TMT values (not ranks)
dfRAW=read.delim("ccle176.mtCN_and_TMT.raw_vals.complete_obs.txt",stringsAsFactors=FALSE)
# make Gene_Symbol the row names and then remove this as a column
rownames(dfRAW)=dfRAW$GeneSym
dfRAW=dfRAW[,2:ncol(dfRAW)]
# create a transposed matrix with genes (and mtCN) as columns and cell lines as rows
dfRAWt=data.frame(t(dfRAW))
# create ranks from raw vals
dfRANK=data.frame(apply(dfRAWt,2,rank,ties.method="min"))
modelsRANK <- regsubsets(mtCN~., data = dfRANK, nvmax = 4, method = "forward",really.big=TRUE)
coef(modelsRANK,1:4)