CpGassoc-package: Association Between Methylation and a Phenotype of Interest

In CpGassoc: Association Between Methylation and a Phenotype of Interest

Description

Is designed to test for association between methylation at CpG sites across the genome and a phenotype of interest, adjusting for any relevant covariates. The package can perform standard analyses of large datasets very quickly with no need to impute the data. It can also handle mixed effects models with chip or batch entering the model as a random intercept. Also includes tools to apply quality control filters, perform permutation tests, and create QQ plots, manhattan plots, and scatterplots for individual CpG sites.

There is a tutorial for CpGassoc available at the website <genetics.emory.edu/conneely/>

Details

Package: CpGassoc Type: Package Title: Association between Methylation and a phenotype of interest Version: 2.60 Date: 2017-05-30 Author: Barfield, R., Conneely, K., Kilaru,V Maintainer: R Barfield <[e-mail]> Description: CpGassoc is designed to test for association between methylation at CpG sites across the genome and a phenotype of interest, adjusting for any relevant covariates. The package can perform standard analyses of large datasets very quickly with no need to impute the data. It can also handle mixed effects models with chip or batch entering the model as a random intercept. CpGassoc also includes tools to apply quality control filters, perform permutation tests, and create QQ plots, manhattan plots, and scatterplots for individual CpG sites. Depends:nlme,methods Suggests:qvalue License: GPL (>= 2)

CpGassoc is a suite of R functions designed to perform flexible analyses of methylation array data. The two main functions are cpg.assoc and cpg.perm. cpg.assoc will perform an association test between the CpG sites and the phenotype of interest. Covariates can be added to the model, and can be continuous or categorical in nature. cpg.assoc allows users to set their own false discovery rate threshold, to transform the beta values to log(beta/(1-beta)), and to subset if required. cpg.assoc can also fit a linear mixed effects model with a single random effect to control for possible technical difference due to batch or chip. cpg.assoc uses the Holm method to determine significance. The user can also specify an FDR method to determine significance based on the function p.adjust or qvalue. cpg.perm performs the same tasks as cpg.assoc followed by a permutation test on the data, repeating the analysis multiple times after randomly permuting the main phenotype of interest. The user can specify the seed and the number of permutations. If over one hundred permutations are performed QQ plots can be created with empirical confidence intervals based on the permuted t-statistics. For more information see plot.cpg.perm. For more information on how to perform cpg.assoc or cpg.perm see their corresponding help pages. CpGassoc can also perform quality control (see cpg.qc).

Author(s)

Barfield, R.; Kilaru,V.; Conneely, K.
Maintainer: R. Barfield: <[e-mail]>

See Also

cpg.assoc cpg.combine cpg.perm cpg.work plot.cpg scatterplot manhattan plot.cpg.perm cpg.qc

Examples

###NOTE: If you are dealing with large data, do not specify large.data=FALSE. 
###The default option is true
##This will involve partitioning up the data and performing more gc() to clear up space
##Using cpg.assoc:
data(samplecpg,samplepheno,package="CpGassoc")
results<-cpg.assoc(samplecpg,samplepheno$weight,large.data=FALSE)
results

##Using cpg.perm:
Testperm<-cpg.perm(samplecpg[1:200,],samplepheno$weight,data.frame(samplepheno$Dose),
                seed=2314,nperm=10,large.data=FALSE)
Testperm


#For more examples go to those two pages main help pages.

Example output

Loading required package: nlme

The top ten CpG sites were:
     CPG.Labels T.statistic      P.value Holm.sig       FDR   gc.p.value
694      CpG694    3.454271 0.0006456268    FALSE 0.4318310 0.0006456268
293      CpG293    3.412320 0.0007485123    FALSE 0.4318310 0.0007485123
560      CpG560    3.313353 0.0010549618    FALSE 0.4318310 0.0010549618
148      CpG148    3.133454 0.0019286973    FALSE 0.5645412 0.0019286973
998      CpG998   -3.079596 0.0022986204    FALSE 0.5645412 0.0022986204
1059    CpG1059   -2.883525 0.0042668430    FALSE 0.7693539 0.0042668430
1182    CpG1182   -2.819710 0.0051827097    FALSE 0.7693539 0.0051827097
100      CpG100    2.787987 0.0057015107    FALSE 0.7693539 0.0057015107
751      CpG751   -2.759379 0.0062093208    FALSE 0.7693539 0.0062093208
238      CpG238    2.756367 0.0062650966    FALSE 0.7693539 0.0062650966

To access results for all  1228  CpG sites use object$results 
or sort(object)$results to obtain results sorted by p-value.

General info:
  Min.P.Observed Num.Cov fdr.cutoff FDR.method Phenotype chipinfo num.Holm
1   0.0006456268       0       0.05         BH    weight     NULL        0
  num.fdr
1       0

0 sites were found significant by the Holm method
0 sites were found significant by BH method

The beta values were taken from: samplecpg 
Effect sizes and standard error can be accessed using $coefficients
Other attributes are: results, Holm.sig, FDR.sig, info, indep, covariates, chip 
 They can be accessed using the $

The permutation P-values, number of permutations and seed:
  p.value.p p.value.holm p.value.FDR nperm seed
1       0.3            1           1    10 2314

Other information:
  Min.P.Observed Num.Cov fdr.cutoff FDR.method num.real.Holm num.real.fdr
1    0.001757539       1       0.05         BH             0            0

The top ten CpG sites were:
    CPG.Labels T.statistic     P.value Holm.sig       FDR  gc.p.value
148     CpG148    3.161818 0.001757539    FALSE 0.3515077 0.007601939
100     CpG100    2.840484 0.004868057    FALSE 0.4868057 0.016341542
52       CpG52   -2.416828 0.016358966    FALSE 0.7097533 0.040732697
85       CpG85    2.345373 0.019775082    FALSE 0.7097533 0.047009419
3         CpG3   -2.288648 0.022918149    FALSE 0.7097533 0.052558145
24       CpG24   -2.090349 0.037576980    FALSE 0.7097533 0.076452721
72       CpG72   -2.061676 0.040252077    FALSE 0.7097533 0.080551514
70       CpG70   -2.051444 0.041245342    FALSE 0.7097533 0.082057103
178     CpG178   -2.042358 0.042144711    FALSE 0.7097533 0.083413116
153     CpG153   -2.025254 0.043883038    FALSE 0.7097533 0.086015232

To access results for all  200  CpG sites use object$results
 or sort(object)$results to obtain results sorted by p-value.

0 sites were found significant by the Holm method
0 sites were found significant by BH method

The beta values were taken from: samplecpg[1:200, ] 
Other attributes are: permutation.matrix, perm.p.values, gc.permutation.matrix, results, Holm.sig ,
 FDR.sig, info, indep, covariates, chip, coefficients. 
They can be accessed using the $

CpGassoc documentation built on May 1, 2019, 8:24 p.m.