1 Instalation

if (!require("BiocManager"))
  install.packages("BiocManager")
BiocManager::install("glmSparseNet")

2 Required Packages

library(dplyr)
library(ggplot2)
library(survival)
library(loose.rock)
library(futile.logger)
library(curatedTCGAData)
library(TCGAutils)
#
library(glmSparseNet)
#
# Some general options for futile.logger the debugging package
.Last.value <- flog.layout(layout.format('[~l] ~m'))
.Last.value <- loose.rock::show.message(FALSE)
# Setting ggplot2 default theme as minimal
theme_set(ggplot2::theme_minimal())

3 Load data

The data is loaded from an online curated dataset downloaded from TCGA using curatedTCGAData bioconductor package and processed.

To accelerate the process we use a very reduced dataset down to around 100 variables only (genes), which is stored as a data object in this package. However, the procedure to obtain the data manually is described in the following chunk.

prad <- curatedTCGAData(diseaseCode = "PRAD", assays = "RNASeq2GeneNorm", FALSE)

Build the survival data from the clinical columns.

• Selects only primary solid tumour samples
• Merge survival times for patients, which have different columns in case they are alive or dead.
• Build two matrix objects that fit the data xdata and ydata

# keep only solid tumour (code: 01)
prad.primary.solid.tumor <- TCGAutils::splitAssays(prad, '01')

## Selecting 'Primary Solid Tumor' samples

xdata.raw <- t(assay(prad.primary.solid.tumor[[1]]))

# Get survival information
ydata.raw <- colData(prad.primary.solid.tumor) %>% as.data.frame %>% 
  # Find max time between all days (ignoring missings)
  rowwise %>%
  mutate(time = max(days_to_last_followup, days_to_death, na.rm = TRUE)) %>%
  # Keep only survival variables and codes
  select(patientID, status = vital_status, time) %>% 
  # Discard individuals with survival time less or equal to 0
  filter(!is.na(time) & time > 0) %>% as.data.frame

# Set index as the patientID
rownames(ydata.raw) <- ydata.raw$patientID

# keep only features that have standard deviation > 0
xdata.raw  <- xdata.raw[TCGAbarcode(rownames(xdata.raw)) %in% 
                          rownames(ydata.raw),]
xdata.raw  <- xdata.raw %>% 
  { (apply(., 2, sd) != 0) } %>% 
  { xdata.raw[, .] } %>% 
  scale

# Order ydata the same as assay
ydata.raw    <- ydata.raw[TCGAbarcode(rownames(xdata.raw)), ]

set.seed(params$seed)
small.subset <- c(geneNames(c('ENSG00000103091', 'ENSG00000064787', 
                              'ENSG00000119915', 'ENSG00000120158', 
                              'ENSG00000114491', 'ENSG00000204176', 
                              'ENSG00000138399'))$external_gene_name, 
                  sample(colnames(xdata.raw), 100)) %>% unique %>% sort

xdata <- xdata.raw[, small.subset[small.subset %in% colnames(xdata.raw)]]
ydata <- ydata.raw %>% select(time, status)

4 Fit models

Fit model model penalizing by the hubs using the cross-validation function by cv.glmHub.

set.seed(params$seed)
fitted <- cv.glmHub(xdata, Surv(ydata$time, ydata$status),
                    family  = 'cox',
                    nlambda = 1000,
                    network = 'correlation', 
                    network.options = networkOptions(cutoff = .6, 
                                                     min.degree = .2))

5 Results of Cross Validation

Shows the results of 100 different parameters used to find the optimal value in 10-fold cross-validation. The two vertical dotted lines represent the best model and a model with less variables selected (genes), but within a standard error distance from the best.

plot(fitted)

coefs.v <- coef(fitted, s = 'lambda.min')[,1] %>% { .[. != 0]}
coefs.v %>% { 
  data.frame(ensembl.id  = names(.), 
             gene.name   = geneNames(names(.))$external_gene_name, 
             coefficient = .,
             stringsAsFactors = FALSE)
  } %>%
  arrange(gene.name) %>%
  knitr::kable()

geneNames(names(coefs.v)) %>% { hallmarks(.$external_gene_name)$heatmap }

separate2GroupsCox(as.vector(coefs.v), 
                   xdata[, names(coefs.v)], 
                   ydata, 
                   plot.title = 'Full dataset', legend.outside = FALSE)

## $pvalue
## [1] 0.06897084
## 
## $plot

## 
## $km
## Call: survfit(formula = survival::Surv(time, status) ~ group, data = prognostic.index.df)
## 
##             n events median 0.95LCL 0.95UCL
## Low risk  249      3     NA    3502      NA
## High risk 248      7     NA    3467      NA