| Title: | Significant Interval Discovery with Categorical Covariates |
| Version: | 0.2.7 |
| Author: | Felipe Llinares Lopez, Dean Bodenham |
| Maintainer: | Dean Bodenham <deanbodenhambsse@gmail.com> |
| Description: | A method which uses the Cochran-Mantel-Haenszel test with significant pattern mining to detect intervals in binary genotype data which are significantly associated with a particular phenotype, while accounting for categorical covariates. |
| Depends: | R (≥ 3.3.0), bindata |
| License: | GPL-2 |GPL-3 |
| LinkingTo: | Rcpp |
| Imports: | Rcpp |
| Suggests: | testthat |
| RoxygenNote: | 5.0.1 |
| SystemRequirements: | C++11 |
| NeedsCompilation: | yes |
| Packaged: | 2016-09-13 16:11:45 UTC; dean |
| Repository: | CRAN |
| Date/Publication: | 2016-09-13 21:19:17 |
Demo of fastcmh
Description
This function runs a demo for fastcmh, by first creating a sample data setand then running fastcmh on this data set.
Usage
demofastcmh(saveToFolder = FALSE, folder = NULL)Arguments
saveToFolder | A flag indicating whether or not the data files createdfor the demo should be saved to file. The default is |
folder | The folder in which the data for the demo will be saved.Default is the current directory, |
Details
This function will first create a sample data set infolder/data,and will then runrunfastcmh on this data set, before saving theeach step showing the R code that can be used to do the step, then runningthat R code, and then waiting for the user to press enter before movingonto the next step. IfsaveToFolder=FALSE, (default) then no filesare saved and all the results are kept in memory.
See Also
Examples
demofastcmh()Create sample data for fastcmh
Description
This function creates sample data for use with therunfastcmh method.
Usage
makefastcmhdata(folder = "./", xfilename = "data.txt", yfilename = "label.txt", covfilename = "cov.txt", K = 2, L = 1000, n = 200, noiseP = 0.3, corruptP = 0.05, rho = 0.8, tau1 = 100, taulength1 = 4, tau2 = 200, taulength2 = 4, seednum = 2, truetaufilename = "truetau.txt", showOutput = FALSE, saveToList = FALSE)Arguments
folder | The folder in which the data will be saved. Default iscurrent directory |
xfilename | The name of the data file. Default is |
yfilename | The name of the label file. Default is |
covfilename | The name of the file containing the covariate categories. This file actually just contains |
K | The number of covariates (a positive integer). Default is |
L | The number of features (length of each sequence). Default is |
n | The number of samples (cases and controls combined). Default is |
noiseP | The background noise in the data (as a probability of 0/1being flipped). Default is |
corruptP | The probability of data corruption: each bit hasprobability |
rho | The strength of the confounding in the confounded interval (asa probability). Default is |
tau1 | The location of the significant interval (starting point).Default value is |
taulength1 | The length of the significant interval. Default value is |
tau2 | The location of the confounded significant interval (startingpoint). Default value is |
taulength2 | The length of the confounded significant interval.Default value is |
seednum | The seed used for generating the data. Default value is |
truetaufilename | The file where the location of the true significantintervals are saved (as opposed to the detected significant intervals).Default is |
showOutput | Flag to decide whether or not to show output, where filesare created, their names, etc. Default is |
saveToList | Flag to decide whether or not to save data to the folder,or to return (output) the data as a list. By default, |
See Also
Examples
#make a small sample data set, using the default parametersmylist <- makefastcmhdata(showOutput=TRUE, saveToList=TRUE)#make a very small sample data setmylist <- makefastcmhdata(n=20, L=10, tau1=2, taulength1=2, tau2=6, taulength2=2, saveToList=TRUE)Run the fastcmh algorithm
Description
This function runs the FastCMH algorithm on a particular data set.
Usage
runfastcmh(folder = NULL, data = NULL, label = NULL, cov = NULL, alpha = 0.05, Lmax = 0, showProcessing = FALSE, saveAllPvals = FALSE, doFDR = FALSE, useDependenceFDR = FALSE, saveToFile = FALSE, saveFilename = "fastcmhresults.RData", saveFolder = NULL)Arguments
folder | The folder in which the data is saved. If the any of |
data | The filename for the data file. Default is |
label | The filename for the phenotype label file. Default is |
cov | The filename for the covariate label file. Default is |
alpha | The value of the FWER; must be a number between 0 and 1.Default is |
Lmax | The maximum length of significant intervals which isconsidered. Must be a non-negative integer. For example, |
showProcessing | A flag which will turn printing to screen on/off.Default is |
saveAllPvals | A flag which controls whether or not all the intervals(less than minimum attainable pvalue) will be returned. Default is |
doFDR | A flag which controls whether or not Gilbert's Tarone FDRprocedure (while accounting for positive regression dependence) isperformed. Default is |
useDependenceFDR | A flag which controls whether or not Gilbert'sTarone FDR procedure uses the dependent formulation by Benjamini andYekutieli (2001), which further adjusts alpha by dividing by the harmonicmean. This flag is only used if |
saveToFile | A flag which controls whether or not the results aresaved to file. By default, |
saveFilename | A string which gives the filename to which the outputis saved (needs to have |
saveFolder | A string which gives the path to which the output willbe saved (needs to have |
Details
This function runs the FastCMH algorithm on a particular data set inorder to discover intervals that are statistically significantlyassociated with a particular label, while accounting for categoricalcovariates.
The user must either supply the folder, which contains files named"data.txt","label.txt" and"cov.txt", or thenon-default filenames must be specified individually. See the descriptions of argumentsdata,label andcov to see the format ofthe input files, or make a small sample data file using themakefastcmhdata function.By default, filtered results are provided. The user also has the optionof using an FDR procedure rather than the standard FWER-preservingprocedure.
Value
runfastcmh will return a list ifsaveToFile=FALSE (defaultsetting), otherwise it will save the list in an .RData file. The fieldsof the list are:
siga dataframe listing the significant intervals, afterfilterting. Columns
start,endandpvalueindicatethe start and end points of the interval (inclusive), and thep-value for that interval.unfiltereda dataframe listing all the significant intervalsbefore filtering. The filtering compares the overlapping intervals andreturns the interval with the smallest p-value in each cluster ofoverlapping intervals. Dataframe has has structure as
sig.fdr(if doFDR==TRUE) significant intervals using Gilbert'sFDR-Tarone procedure, after filtering. Dataframe has same structure as
sig.unfilteredFdr(if doFDR==TRUE) a dataframe listing all the significant intervals before filtering. See description of
unfiltered.allTestablle(if saveAllPvals==TRUE) a dataframe listing allthe testable intervals, many of which will not be significant. Dataframehas same structure as
sig.histObsTogether with histFreq gives a histogram of maximumattainable CMH statistics.
histFreqHistogram of maximum attainable CMH statistics (onlyreliable in the testable range).
summarya character string summarising the results. Use
cat(...$summary)to print the results with the correctindentation/new lines.timinga list containing (i)
details, a characterstring summarising the runtime values for the experiment - usecat(...$timing$details)for correct indentation, etc.(ii)exec, the total execution time. (iii)init, the timeto initialise the objects. (iv)fileIO, the time to read the inputfiles. (v)compSigThresh, the time to compute the significancethreshold. (vi)compSigInt, the time to compute the significantintervals.
Author(s)
Felipe Llinares Lopez, Dean Bodenham
See Also
References
Gilbert, P. B. (2005)A modified false discovery ratemultipl-comparisons procedure for discrete data, applied to humanimmunodeficiency virus genetics. Journal of the Royal StatisticalSociety: Series C (Applied Statistics), 54(1), 143-158.
Benjamini, Y., Yekutieli, D. (2001).The control of the falsediscovery rate in multiple testing under dependency.Annals of Statistics, 29(4), 1165-1188.
Examples
#Example with default naming convention used for data, label and cov files# Note: using "/data/" as the argument for folder# accesses the data/ directory in the fastcmh package foldermylist <- runfastcmh("/data/")#Example where the progress will be shownmylist <- runfastcmh(folder="/data/", showProcessing=TRUE)#Example where many parameters are specifiedmylist <- runfastcmh(folder="/data/", data="data2.txt", alpha=0.01, Lmax=7)#Example where Gilbert's Tarone-FDR procedure is usedmylist <- runfastcmh("/data/", doFDR=TRUE)#Example where FDR procedure takes some dependence structures into accountmylist <- runfastcmh("/data/", doFDR=TRUE, useDependenceFDR=TRUE)