Correlation between mutation rate and clinical features
Pheochromocytoma and Paraganglioma (Primary solid tumor)
17 October 2014  |  analyses__2014_10_17
Maintainer Information
Citation Information
Maintained by Juok Cho (Broad Institute)
Cite as Broad Institute TCGA Genome Data Analysis Center (2014): Correlation between mutation rate and clinical features. Broad Institute of MIT and Harvard. doi:10.7908/C1GB22Z0
Overview
Introduction

This pipeline uses various statistical tests to identify selected clinical features related to mutation rate.

Summary

Testing the association between 2 variables and 6 clinical features across 132 samples, statistically thresholded by P value < 0.05 and Q value < 0.3, 3 clinical features related to at least one variables.

  • 1 variable correlated to 'Time to Death'.

    • MUTATIONRATE_SILENT

  • 2 variables correlated to 'AGE'.

    • MUTATIONRATE_NONSYNONYMOUS ,  MUTATIONRATE_SILENT

  • 2 variables correlated to 'AGE_mutation.rate'.

    • MUTATIONRATE_SILENT ,  MUTATIONRATE_NONSYNONYMOUS

  • No variables correlated to 'GENDER', 'RACE', and 'ETHNICITY'.

Results
Overview of the results

Complete statistical result table is provided in Supplement Table 1

Table 1.  Get Full Table This table shows the clinical features, statistical methods used, and the number of variables that are significantly associated with each clinical feature at P value < 0.05 and Q value < 0.3.

Clinical feature Statistical test Significant variables Associated with                 Associated with
Time to Death Cox regression test N=1 shorter survival N=1 longer survival N=0
AGE Spearman correlation test N=2 older N=2 younger N=0
AGE Linear Regression Analysis N=2        
GENDER Wilcoxon test   N=0        
RACE Kruskal-Wallis test   N=0        
ETHNICITY Wilcoxon test   N=0        
Clinical variable #1: 'Time to Death'

One variable related to 'Time to Death'.

Table S1.  Basic characteristics of clinical feature: 'Time to Death'

Time to Death Duration (Months) 0.1-303.1 (median=16.9)
  censored N = 128
  death N = 4
     
  Significant variables N = 1
  associated with shorter survival 1
  associated with longer survival 0
List of one variable associated with 'Time to Death'

Table S2.  Get Full Table List of one variable significantly associated with 'Time to Death' by Cox regression test

HazardRatio Wald_P Q C_index
MUTATIONRATE_SILENT Inf 0.003769 0.0075 0.813
Clinical variable #2: 'AGE'

2 variables related to 'AGE'.

Table S3.  Basic characteristics of clinical feature: 'AGE'

AGE Mean (SD) 48.8 (16)
  Significant variables N = 2
  pos. correlated 2
  neg. correlated 0
List of 2 variables associated with 'AGE'

Table S4.  Get Full Table List of 2 variables significantly correlated to 'AGE' by Spearman correlation test

SpearmanCorr corrP Q
MUTATIONRATE_NONSYNONYMOUS 0.454 4.555e-08 9.11e-08
MUTATIONRATE_SILENT 0.2421 0.005169 0.00517
Clinical variable #3: 'AGE'

2 variables related to 'AGE'.

Table S5.  Basic characteristics of clinical feature: 'AGE'

AGE Mean (SD) 48.8 (16)
  Significant variables N = 2
List of 2 variables associated with 'AGE'

Table S6.  Get Full Table List of 2 variables significantly correlated to 'AGE' by Linear regression analysis [lm (mutation rate ~ age)]. Compared to a correlation analysis testing for interdependence of the variables, a regression model attempts to describe the dependence of a variable on one (or more) explanatory variables assuming that there is a one-way causal effect from the explanatory variable(s) to the response variable. If 'Residuals vs Fitted' plot (a standard residual plot) shows a random pattern indicating a good fit for a linear model, it explains linear regression relationship between Mutation rate and age factor. Adj.R-squared (= Explained variation / Total variation) indicates regression model's explanatory power.

Adj.R.squared F P Residual.std.err DF coef(intercept) coef.p(intercept)
MUTATIONRATE_SILENT 0.043 6.89 0.00971 8.08e-08 130 1.18e-09 ( 6.59e-08 ) 0.00971 ( 0.00485 )
MUTATIONRATE_NONSYNONYMOUS 0.183 30.4 1.81e-07 1.62e-07 130 4.96e-09 ( 1.66e-07 ) 1.81e-07 ( 0.000442 )
Clinical variable #4: 'GENDER'

No variable related to 'GENDER'.

Table S7.  Basic characteristics of clinical feature: 'GENDER'

GENDER Labels N
  FEMALE 75
  MALE 57
     
  Significant variables N = 0
Clinical variable #5: 'RACE'

No variable related to 'RACE'.

Table S8.  Basic characteristics of clinical feature: 'RACE'

RACE Labels N
  AMERICAN INDIAN OR ALASKA NATIVE 1
  ASIAN 5
  BLACK OR AFRICAN AMERICAN 9
  WHITE 114
     
  Significant variables N = 0
Clinical variable #6: 'ETHNICITY'

No variable related to 'ETHNICITY'.

Table S9.  Basic characteristics of clinical feature: 'ETHNICITY'

ETHNICITY Labels N
  HISPANIC OR LATINO 3
  NOT HISPANIC OR LATINO 101
     
  Significant variables N = 0
Methods & Data
Input
  • Expresson data file = PCPG-TP.patients.counts_and_rates.txt

  • Clinical data file = PCPG-TP.merged_data.txt

  • Number of patients = 132

  • Number of variables = 2

  • Number of clinical features = 6

Survival analysis

For survival clinical features, Wald's test in univariate Cox regression analysis with proportional hazards model (Andersen and Gill 1982) was used to estimate the P values using the 'coxph' function in R. Kaplan-Meier survival curves were plot using the four quartile subgroups of patients based on expression levels

Correlation analysis

For continuous numerical clinical features, Spearman's rank correlation coefficients (Spearman 1904) and two-tailed P values were estimated using 'cor.test' function in R

Student's t-test analysis

For two-class clinical features, two-tailed Student's t test with unequal variance (Lehmann and Romano 2005) was applied to compare the log2-expression levels between the two clinical classes using 't.test' function in R

ANOVA analysis

For multi-class clinical features (ordinal or nominal), one-way analysis of variance (Howell 2002) was applied to compare the log2-expression levels between different clinical classes using 'anova' function in R

Q value calculation

For multiple hypothesis correction, Q value is the False Discovery Rate (FDR) analogue of the P value (Benjamini and Hochberg 1995), defined as the minimum FDR at which the test may be called significant. We used the 'Benjamini and Hochberg' method of 'p.adjust' function in R to convert P values into Q values.

Download Results

In addition to the links below, the full results of the analysis summarized in this report can also be downloaded programmatically using firehose_get, or interactively from either the Broad GDAC website or TCGA Data Coordination Center Portal.

References
[1] Andersen and Gill, Cox's regression model for counting processes, a large sample study, Annals of Statistics 10(4):1100-1120 (1982)
[2] Spearman, C, The proof and measurement of association between two things, Amer. J. Psychol 15:72-101 (1904)
[3] Lehmann and Romano, Testing Statistical Hypotheses (3E ed.), New York: Springer. ISBN 0387988645 (2005)
[4] Howell, D, Statistical Methods for Psychology. (5th ed.), Duxbury Press:324-5 (2002)
[5] Benjamini and Hochberg, Controlling the false discovery rate: a practical and powerful approach to multiple testing, Journal of the Royal Statistical Society Series B 59:289-300 (1995)