title: “Basic Usage”

author: “Dominic Pearce”

date: “2017-08-11”

header-includes:

-

output: github_document

vignette: >

%

%

%

 

Quick example using NKI breast cancer data

note: this data is pre-subsetted to only include patients with complete distant metastasis information (e.dmfs & t.dmfs)

 

library(survivALL)
library(Biobase)
library(ggplot2)

data(nki_subset)

 

Example : Plotting

We use a continuous measure, here a vector of expression, to re-order our survival data and then compute hazard ratios and pvalues for all points of separation

xpr_vec <- exprs(nki_subset)["NM_001758", ] #expression vector for CCND1 (a marker of proliferation)

plotALL(
        measure = xpr_vec, #expression data
        srv = pData(nki_subset), #survival information
        time = "t.dmfs", #time-to-outcome
        event = "e.dmfs", #outcome type
        bs_dfr = c(), #thresholding data would go here
        measure_name = "CCND1", #our gene's name
        plot_range = "auto", #we can specify the plots y-axis range here. This makes comparing plots easier
        scale_upper = "auto", #similarly we can specify an upper limit to our pvalue colour scale
        title = "CCND1 prognostic capacity in a mixed\npopulation of invasive breast cancer samples", #plot title
        legend = "bottom", #legend position, one of 'top', 'right', 'bottom', 'left' or 'none'
        axis_text = TRUE #axis text can be removed to more easily compare multiple plots side-by-side
        )

 

Note that we can add additional elements using standard ggplot2 syntax. Here we add a horizontal indicator of the most significant point of separation

plotALL(measure = xpr_vec, 
        srv = pData(nki_subset), 
        time = "t.dmfs", 
        event = "e.dmfs", 
        bs_dfr = c(),
        measure_name = "CCND1", 
        title = "CCND1 prognostic capacity in a mixed\npopulation of invasive breast cancer samples") + 
    geom_vline(xintercept = 290, linetype = 5)

Plotting multiple genes simultaneously

We first organise our measure data, our expression vectors for three genes of interest CCND1, FOS and ERBB2 before applying each in a loop, specifying a common and sensible y-axis range. (To choose the limits we produce the plots first, select a rational range by eye and then recompute with the newly specified limits). We then combine the figures using the cowplot::plot_grid() function.

geneset <- data.frame(refseq_id = c("NM_001758", "NM_005252", "NM_004448"), hgnc_id = c("CCND1", "FOS", "ERBB2"), stringsAsFactors = FALSE)

xpr_lst <- lapply(geneset$refseq_id, function(id){
                exprs(nki_subset)[id,]
        })
names(xpr_lst) <- geneset$hgnc_id

plot_lst <- lapply(geneset$hgnc_id, function(id){
                       plotALL(
                               measure = xpr_lst[[id]], #expression data
                               srv = pData(nki_subset), #survival information
                               time = "t.dmfs", #time-to-outcome
                               event = "e.dmfs", #outcome type
                               bs_dfr = c(), #thresholding data 
                               measure_name = id, #our gene's name
                               plot_range = c(-2.5, 2.5), #we can specify the plots y-axis range here. This makes comparing plots easier
                               scale_upper = "auto", #similarly we can specify an upper limit to our pvalue colour scale
                               title = id, #plot title
                               legend = "bottom", #legend position, one of 'top', 'right', 'bottom', 'left' or 'none'
                               axis_text = TRUE #axis text can be removed to more easily compare multiple plots side-by-side
                               )
        })

cowplot::plot_grid(plotlist = plot_lst, nrow = 1)

Example : Returning a dataframe

Alternatively, we can return only the computed statistics as a dataframe for further calculations, comparisons and manipulations

survivall_out <- survivALL(
                           measure = xpr_vec, #expression data
                           srv = pData(nki_subset), #survival information
                           time = "t.dmfs", #time-to-outcome
                           event = "e.dmfs", #outcome type
                           bs_dfr = c(), #thresholding data
                           measure_name = "CCND1", #our gene's name
                           n_sd = 1.96 #number of standard deviations that define the threshold width - 1.96 represents 95% of random associations fall within the thresholds
                           )

head(survivall_out)

Table continues below

index	samples	event_time	event	measure	HR	p	log10_p
1	NKI_270	94	FALSE	-0.982	NA	0	Inf
2	NKI_106	701	TRUE	-0.934	-3.364	0.09253	NA
3	NKI_184	442	TRUE	-0.91	-3.816	0.009459	2.024
4	NKI_197	4035	FALSE	-0.909	-1.646	0.1773	NA
5	NKI_226	972	FALSE	-0.868	-1.287	0.2751	NA
6	NKI_215	3781	FALSE	-0.832	-0.6191	0.5742	NA

name	mean	sdplus	sdmin	threshold_residuals	dsr	most_dsr	clsf
CCND1	0	0	0	NA	0	FALSE	0
CCND1	0	0	0	3.364	0	FALSE	1
CCND1	0	0	0	3.816	0	FALSE	1
CCND1	0	0	0	1.646	0	FALSE	1
CCND1	0	0	0	1.287	0	FALSE	1
CCND1	0	0	0	0.6191	0	FALSE	1

Calculating multiple genes

We can return the results for multiple genes as a single dataframe simply by row-binding the results. Organised in this way we can plot multiple hazard ratio distributions as a single figure

survivall_lst <- lapply(geneset$hgnc_id, function(id){
                            survivALL(
                                      measure = xpr_lst[[id]], #expression data
                                      srv = pData(nki_subset), #survival information
                                      time = "t.dmfs", #time-to-outcome
                                      event = "e.dmfs", #outcome type
                                      bs_dfr = c(), #thresholding data
                                      measure_name = id, #our gene's name
                                      n_sd = 1.96 #number of standard deviations that define the threshold width - 1.96 represents 95% of random associations fall within the thresholds
                                      )
                           })

survivall_dfr <- do.call(rbind, survivall_lst)

ggplot(survivall_dfr, aes(x = index, y = HR, colour = name)) + 
    geom_hline(yintercept = 0, linetype = 3) + 
    geom_point() + 
    ylim(-2.5, 2.5) + 
    ggthemes::theme_pander()