library(gnomeR)

Introduction

In this article we will cover functions integrated in gnomeR to analyze the outputed processed data of the binmat() function. We include in gnomeR the following functions for analysis:

gen.summary() which test for association between either categorical or continous outcomes and the set of genomic features from the binmat() output matrix. Returns a table with the distribution of the genomic features along the outcome specified, using adapted tests and adjustements for multiple testing to report significance. We further summarize the results using adequate plots.
uni.cox() which performs univariate assocition analysis between a time-dependent outcome and the set of genomic features from the binmat() output matrix. We perform this using Cox’s proportional hazard models with multiple test adjustements and summary plots.

Categorical and continous outcomes

In this first section we focus on the gen.summary() function, it takes the following arguments:

gen.dat a genomic dataframe outputed from binmat().
outcome a leveled vector of length equal to the number of rows in gen.dat.
filter a numeric value between 0 and 1 (1 not included) that is the lower bound for the proportion of patients having a genetic event. All features with an event rate lower than that value will be removed. Default is 0 (all features included).
paired Boolean if the data are paired. Default is FALSE.
cont Should the outcome be treated as a continuous value. Default is FALSE treated as categorical.
rank Should the table returned be ordered by Pvalue. Boolean, default is TRUE.

In the following analyses we will be using the following subset of samples:

set.seed(123)
samples <- as.character(unique(mut$Tumor_Sample_Barcode))[sample(1:length(unique(mut$Tumor_Sample_Barcode)), 100, replace=FALSE)]
df <- binmat(patients = samples,maf = mut, fusion = fusion, cna = cna, cna.binary = FALSE)

Categorical

For this example we will be using the recorded sample type of the samples we randomly selected above. Thus looking for differences between primary and metastatic samples:

outcome <- as.character(clin.sample$Sample.Type[match(samples,clin.sample$Sample.Identifier)])
test <- gen.summary(gen.dat = df,
                outcome = outcome,
                filter = 0.05,paired = FALSE,
                cont = FALSE,rank = TRUE)

Using the Fisher’s exact test the gen.summary() function returns the following:

fits a summary table of the distribution of features across outcomes and the test performed

kable(test$fits[1:10,])

	Feature	Overall	Metastasis(N=55)	Primary(N=45)	OddsRatio	Pvalue	FDR	Lower	Upper
MLL2	MLL2	5%	9.09%	0%	0	6.24e-02	1.00e+00	0	1.29
CREBBP	CREBBP	7%	10.91%	2.22%	0.19	1.25e-01	1.00e+00	0	1.64
MLL3	MLL3	6%	9.09%	2.22%	0.23	2.19e-01	1.00e+00	0	2.17
CDH1	CDH1	6%	9.09%	2.22%	0.23	2.19e-01	1.00e+00	0	2.17
MYC.cna	MYC.cna	12%	18.18%	4.44%	0.23	2.19e-01	1.00e+00	0	2.17
EGFR	EGFR	9%	5.45%	13.33%	2.64	2.92e-01	1.00e+00	0.52	17.34
KEAP1	KEAP1	5%	7.27%	2.22%	0.29	3.75e-01	1.00e+00	0.01	3.1
AR	AR	5%	7.27%	2.22%	0.29	3.75e-01	1.00e+00	0.01	3.1
SMARCA4	SMARCA4	5%	7.27%	2.22%	0.29	3.75e-01	1.00e+00	0.01	3.1
CCND1.cna	CCND1.cna	10%	14.55%	4.44%	0.29	3.75e-01	1.00e+00	0.01	3.1

forest.plot a forest plot of the most significant features

test$forest.plot

vPlot a volcano plot summarising the odds ratios and pvalues of the the features included in the test

test$vPlot

Continous

In this section we show to use the gen.summary() function with a continuous outcome. Given the lack of continuous variables in the IMPACT clinical set we show an example here with a simulate continuous gaussian outcome. We perform this using linear regression.

set.seed(1)
outcome <-  rnorm(n = nrow(df))
tab.out <- gen.summary(gen.dat = df,
                   outcome = outcome,
                   filter = 0.05,paired = FALSE,
                   cont = TRUE,rank = TRUE)

This returns the following:

fits a summary of the univariate fits that were performed

kable(tab.out$fits[1:10,])

	Estimate	SD	Pvalue	MutationFreq	FDR
EPHA5	0.7144603	0.4078836	0.0829663	0.05	0.9710584
CDKN2B.cna	0.3244820	0.1872198	0.0862125	-0.12	0.9710584
AR	0.6954129	0.4082194	0.0916387	0.05	0.9710584
PTPRT	0.6081550	0.3751380	0.1081987	0.06	0.9710584
RB1	-0.5336722	0.3762929	0.1592940	0.06	0.9710584
TSC1	0.4379804	0.4118499	0.2901928	0.05	0.9710584
TP53	0.1897100	0.1850001	0.3076714	0.38	0.9710584
DOT1L	-0.3573356	0.4126437	0.3886236	0.05	0.9710584
CCND1.cna	-0.1742028	0.2063608	0.4006325	0.10	0.9710584
FGF19.cna	-0.1742028	0.2063608	0.4006325	0.10	0.9710584

vPlot a volcano plot summarising the coefficient and pvalues of the the features included in the test

tab.out$vPlot

Survival outcome

The univariate survival analysis can be performed using the uni.cox() function, it takes the following arguments:

X a matrix/surv.datframe of genomic features, continuous or binary (note cannot handle categorical surv.dat for the moment).
surv.dat a survival dataframe containing the survival information. This can be made of 2 or 3 columns. 1 or 2 for time, and one for status (where 1 is event and 0 is no event).
surv.formula a survival formula with names matching those in surv.dat eg: Surv(time,status)~.
filter a numeric value between 0 and 1 (1 not included) that is the lower bound for the proportion of patients having a genetic event. All features with an event rate lower than that value will be removed. Default is 0 (all features included).
genes a character vector of gene names that will be the only ones to be kept. Default is NULL, all genes are used.

Here we show an example using the survival data included in the IMPACT clinical dataset:

surv.dat <- clin.patients %>%
  filter(X.Patient.Identifier %in%
           abbreviate(samples,strict = TRUE, minlength = 9)) %>%
  select(X.Patient.Identifier,Overall.Survival..Months.,
         Overall.Survival.Status) %>%
  rename(DMPID = X.Patient.Identifier,
         time = Overall.Survival..Months.,
         status = Overall.Survival.Status) %>%
  mutate(time = as.numeric(as.character(time)),
         status = ifelse(status == "LIVING",0,1)) %>%
  filter(!is.na(time))
X <- df[match(surv.dat$DMPID,
                   abbreviate(rownames(df),strict = TRUE, minlength = 9)),]
test <- uni.cox(X = X, surv.dat = surv.dat,
        surv.formula = Surv(time,status)~.,filter = 0.1)