loessByCluster returning error in smooth function in bumphunter
1
0
Entering edit mode
Jiayi Sun ▴ 40
@jiayi-sun-4938
Last seen 5.1 years ago
Hello, I have been trying to use bumphunter on my 450k data and I have been running into some problems with the smooth function using the option of loessByCluster. I have tired both other smoothing functions runmedByCluster and locfitByCluster and these perform correctly however I get this error when trying to use loessByCluster. Error in { : task 1 failed - "NA/NaN/Inf in foreign function call (arg 1)" I have called bumphunter with the following function: ### Run all chromosomes together chr_locs <- ordered by chromosome genomic location chr_chr <- chromosome information chr_clusters <- clusterMaker(chr_chr, chr_locs, maxGap = 400) logtrans <- log transformed beta-values chr_bumphunter <- bumphunter(logtrans, design matrix with first column of "1" and second column with membership comparing 6 (0) samples to 89 samples (1), chr_chr, chr_locs, chr_clusters, smoothFunction = loessByCluster, smooth = TRUE) This error always occurs when I am trying to do the permutations, never on the normal data. This is the output from bumphunter. bumphunterEngine: Using a single core (backend: doSEQ, version: 1.4.2) bumphunterEngine: Computing coefficients. bumphunterEngine: Smoothing coefficients. bumphunterEngine cutoff: 1.123 bumphunterEngine: Finding regions. bumphunterEngine: Performing 100 permutations. bumphunterEngine: Computing marginal permutation p-values. bumphunterEngine: Smoothing permutation coefficients. After which I see the error above. I have tired checking the logtransformed beta-values to see if I have any NAs or 1 but there aren't any. Any idea what this error is coming from? I do have some very, very small numbers approaching zero but I have tired even to change these to 0.001 and it does not fix the error. Furthermore, it's only in the permutation portion where this error occurs so I thought this might not be the problem. Any ideas what's going on? Thanks! [[alternative HTML version deleted]]
bumphunter bumphunter • 1.2k views
0
Entering edit mode
Jiayi Sun ▴ 40
@jiayi-sun-4938
Last seen 5.1 years ago
Sorry, I forgot to include the traceback in the last message. Here it is. 10: stop(simpleError(msg, call = expr)) 9: e$fun(obj, substitute(ex), parent.frame(), e$data) 8: list(args = IndexesChunks(.doRNG.stream = c(407L, -129130598L, 1725800531L, 1089509752L, -772932327L, 780595718L, 241179855L )), argnames = c("idx", ".doRNG.stream"), evalenv = <environment>, specified = character(0), combineInfo = list(fun = function (a, ...) c(a, list(...)), in.order = TRUE, has.init = TRUE, init = list(), final = NULL, multi.combine = TRUE, max.combine = 100), errorHandling = "stop", packages = c("bumphunter", "doRNG" ), export = NULL, noexport = NULL, options = list(), verbose = FALSE) %dopar% { { rngtools::RNGseed(.doRNG.stream) } { idx <- get(foreach_argname) smoothFunction(y = y[idx, ], x = x[idx], cluster = cluster[idx], weights = weights[idx, ], verbose = verbose, ...) } } 7: do.call("%dopar%", list(obj, ex), envir = parent.frame()) 6: foreach_object %dorng% { idx <- get(foreach_argname) smoothFunction(y = y[idx, ], x = x[idx], cluster = cluster[idx], weights = weights[idx, ], verbose = verbose, ...) } 5: smoother(y = permRawBeta, x = pos, cluster = cluster, weights = weights, smoothFunction = smoothFunction, verbose = subverbose) 4: bumphunterEngine(object, design = design, chr = chr, pos, cluster = cluster, coef = coef, cutoff = cutoff, cutoffQ = cutoffQ, maxGap = maxGap, smooth = smooth, smoothFunction = smoothFunction, useWeights = useWeights, B = B, verbose = verbose, ...) 3: .local(object, ...) 2: bumphunter(logtrans, matrix(c(rep(1, ncol(final_Avg_beta_disc_NMF)), as.numeric(factor(fullclinical[match(colnames(final_Avg_beta_disc_NMF) , fullclinical$OffTargetName), ]$NMF_2)) - 1), nrow = ncol(final_Avg_beta_disc_NMF)), chr_chr, chr_locs$MAPINFO, chr_clusters, smoothFunction = loessByCluster, smooth = TRUE) 1: bumphunter(logtrans, matrix(c(rep(1, ncol(final_Avg_beta_disc_NMF)), as.numeric(factor(fullclinical[match(colnames(final_Avg_beta_disc_NMF) , fullclinical$OffTargetName), ]$NMF_2)) - 1), nrow = ncol(final_Avg_beta_disc_NMF)), chr_chr, chr_locs$MAPINFO, chr_clusters, smoothFunction = loessByCluster, smooth = TRUE) On Tue, Jun 3, 2014 at 1:37 PM, Jiayi Sun <monikasun88@gmail.com> wrote: > Hello, > > I have been trying to use bumphunter on my 450k data and I have been > running into some problems with the smooth function using the option of > loessByCluster. I have tired both other smoothing functions > runmedByCluster and locfitByCluster and these perform correctly however I > get this error when trying to use loessByCluster. > > Error in { : > task 1 failed - "NA/NaN/Inf in foreign function call (arg 1)" > > I have called bumphunter with the following function: > > ### Run all chromosomes together > > chr_locs <- ordered by chromosome genomic location > chr_chr <- chromosome information > chr_clusters <- clusterMaker(chr_chr, chr_locs, maxGap = 400) > > logtrans <- log transformed beta-values > > chr_bumphunter <- bumphunter(logtrans, > design matrix with first column of "1" and second column with membership > comparing 6 (0) samples to 89 samples (1), > chr_chr, chr_locs, chr_clusters, smoothFunction = loessByCluster, smooth > = TRUE) > > This error always occurs when I am trying to do the permutations, never on > the normal data. This is the output from bumphunter. > > bumphunterEngine: Using a single core (backend: doSEQ, version: 1.4.2) > bumphunterEngine: Computing coefficients. > bumphunterEngine: Smoothing coefficients. > bumphunterEngine cutoff: 1.123 > bumphunterEngine: Finding regions. > bumphunterEngine: Performing 100 permutations. > bumphunterEngine: Computing marginal permutation p-values. > bumphunterEngine: Smoothing permutation coefficients. > > After which I see the error above. I have tired checking the > logtransformed beta-values to see if I have any NAs or 1 but there aren't > any. Any idea what this error is coming from? I do have some very, very > small numbers approaching zero but I have tired even to change these to > 0.001 and it does not fix the error. Furthermore, it's only in the > permutation portion where this error occurs so I thought this might not be > the problem. > > Any ideas what's going on? > > Thanks! > > > -- Jiayi Monika Sun Rice University, Bioengineering '10 Baylor College of Medicine Graduate School, SCBMB jmsun@bcm.edu [[alternative HTML version deleted]]