The ramr User’s Guide

Oleksii Nikolaienko

26 October, 2021

Abstract

A comprehensive guide to using the ramr package for detection of rare aberrantly methylated regions.

Introduction

ramr is an R package for detection of low-frequency aberrant methylation events in large data sets obtained by methylation profiling using array or high-throughput bisulfite sequencing. In addition, package provides functions to visualize found aberrantly methylated regions (AMRs), to generate sets of all possible regions to be used as reference sets for enrichment analysis, and to generate biologically relevant test data sets for performance evaluation of AMR/DMR search algorithms.

Current Features

Reading data

ramr methods operate on objects of the class GRanges. The input object for AMR search must in addition contain metadata columns with sample beta values. A typical input object looks like this:

GRanges object with 383788 ranges and 845 metadata columns:
             seqnames    ranges strand |         GSM1235534         GSM1235535         GSM1235536 ...
                <Rle> <IRanges>  <Rle> |          <numeric>          <numeric>          <numeric> ...
  cg13869341     chr1     15865      * |  0.801634776091808  0.846486905008704   0.86732154737116 ...
  cg24669183     chr1    534242      * |  0.834138820071765  0.861974610731835  0.832557979806823 ...
  cg15560884     chr1    710097      * |  0.711275180750356   0.70461945838556  0.699487225634589 ...
  cg01014490     chr1    714177      * | 0.0769098196182058 0.0569443780518647 0.0623154673389864 ...
  cg17505339     chr1    720865      * |  0.876413362222415  0.885593263385521  0.877944732153869 ...
         ...      ...       ...    ... .                ...                ...                ... ...
  cg05615487    chr22  51176407      * |   0.84904178467798  0.836538383875097   0.81568519870099 ...
  cg22122449    chr22  51176711      * |  0.882444486059592  0.870804215405886  0.859269224277308 ...
  cg08423507    chr22  51177982      * |  0.886406345093286  0.882430879852752  0.887241923657461 ...
  cg19565306    chr22  51222011      * | 0.0719084295670266 0.0845209871264646 0.0689074604483659 ...
  cg09226288    chr22  51225561      * |  0.724145303755024  0.696281176451351  0.711459675603635 ...

ramr package is supplied with a sample data, which was simulated using GSE51032 data set as described in the ramr reference paper. Sample data set ramr.data contains beta values for 10000 CpGs and 100 samples (ramr.samples), and carries 6 unique (ramr.tp.unique) and 15 non-unique (ramr.tp.nonunique) true positive AMRs containing at least 10 CpGs with their beta values increased/decreased by 0.5

library(ramr)
#> Loading required package: GenomicRanges
#> Loading required package: stats4
#> Loading required package: BiocGenerics
#> 
#> Attaching package: 'BiocGenerics'
#> The following objects are masked from 'package:stats':
#> 
#>     IQR, mad, sd, var, xtabs
#> The following objects are masked from 'package:base':
#> 
#>     Filter, Find, Map, Position, Reduce, anyDuplicated, append, as.data.frame, basename,
#>     cbind, colnames, dirname, do.call, duplicated, eval, evalq, get, grep, grepl,
#>     intersect, is.unsorted, lapply, mapply, match, mget, order, paste, pmax, pmax.int,
#>     pmin, pmin.int, rank, rbind, rownames, sapply, setdiff, sort, table, tapply, union,
#>     unique, unsplit, which.max, which.min
#> Loading required package: S4Vectors
#> 
#> Attaching package: 'S4Vectors'
#> The following objects are masked from 'package:base':
#> 
#>     I, expand.grid, unname
#> Loading required package: IRanges
#> Loading required package: GenomeInfoDb
#> Loading required package: parallel
#> Loading required package: doParallel
#> Loading required package: foreach
#> Loading required package: iterators
#> Loading required package: doRNG
#> Loading required package: rngtools
data(ramr)

head(ramr.samples)
#> [1] "sample1" "sample2" "sample3" "sample4" "sample5" "sample6"
ramr.data[1:10,ramr.samples[1:3]]
#> GRanges object with 10 ranges and 3 metadata columns:
#>              seqnames    ranges strand |   sample1   sample2   sample3
#>                 <Rle> <IRanges>  <Rle> | <numeric> <numeric> <numeric>
#>   cg13869341     chr1     15865      * |  0.833609  0.847747  0.879959
#>   cg14008030     chr1     18827      * |  0.553312  0.547931  0.609759
#>   cg12045430     chr1     29407      * |  0.186572  0.204322  0.222865
#>   cg20826792     chr1     29425      * |  0.481280  0.439688  0.426783
#>   cg00381604     chr1     29435      * |  0.134126  0.173629  0.167011
#>   cg20253340     chr1     68849      * |  0.500267  0.615868  0.541587
#>   cg21870274     chr1     69591      * |  0.777613  0.771680  0.775057
#>   cg03130891     chr1     91550      * |  0.245783  0.204307  0.223610
#>   cg24335620     chr1    135252      * |  0.777796  0.790842  0.786113
#>   cg16162899     chr1    449076      * |  0.878150  0.860134  0.898071
#>   -------
#>   seqinfo: 24 sequences from hg19 genome; no seqlengths
plotAMR(ramr.data, ramr.samples, ramr.tp.unique[1])
#> [[1]]

plotAMR(ramr.data, ramr.samples, ramr.tp.nonunique[c(1,6,11)])
#> [[1]]

The input (or template) object may be obtained using data from various sources. Here we provide two examples:

Using data from NCBI GEO

The following code pulls (NB: very large) raw files from NCBI GEO database, performes normalization and creates GRanges object for further analysis using ramr (system requirements: 22GB of disk space, 64GB of RAM)

library(minfi)
library(GEOquery)
library(GenomicRanges)
library(IlluminaHumanMethylation450kanno.ilmn12.hg19)

# destination for temporary files
dest.dir <- tempdir()

# downloading and unpacking raw IDAT files
suppl.files <- getGEOSuppFiles("GSE51032", baseDir=dest.dir, makeDirectory=FALSE, filter_regex="RAW")
untar(rownames(suppl.files), exdir=dest.dir, verbose=TRUE)
idat.files  <- list.files(dest.dir, pattern="idat.gz$", full.names=TRUE)
sapply(idat.files, gunzip, overwrite=TRUE)

# reading IDAT files
geo.idat <- read.metharray.exp(dest.dir)
colnames(geo.idat) <- gsub("(GSM\\d+).*", "\\1", colnames(geo.idat))

# processing raw data
genomic.ratio.set <- preprocessQuantile(geo.idat, mergeManifest=TRUE, fixOutliers=TRUE)

# creating the GRanges object with beta values
data.ranges <- granges(genomic.ratio.set)
data.betas  <- getBeta(genomic.ratio.set)
sample.ids  <- colnames(geo.idat)
mcols(data.ranges) <- data.betas

# data.ranges and sample.ids objects are now ready for AMR search using ramr

Using Bismark cytosine report files

library(methylKit)
library(GenomicRanges)

# file.list is a user-defined character vector with full file names of Bismark cytosine report files
file.list

# sample.ids is a user-defined character vector holding sample names
sample.ids

# methylation context string, defines if the reads covering both strands will be merged
context <- "CpG"

# fitting beta distribution (filtering using ramr.method "beta" or "wbeta") requires
# that most of the beta values are not equal to 0 or 1
min.beta <- 0.001
max.beta <- 0.999

# reading and uniting methylation values
meth.data.raw <- methRead(as.list(file.list), as.list(sample.ids), assembly="hg19", header=TRUE,
                          context=context, resolution="base", treatment=rep(0,length(sample.ids)),
                          pipeline="bismarkCytosineReport")
meth.data.utd <- unite(meth.data.raw, destrand=isTRUE(context=="CpG"))

# creating the GRanges object with beta values
data.ranges <- GRanges(meth.data.utd)
data.betas  <- percMethylation(meth.data.utd)/100
data.betas[data.betas<min.beta] <- min.beta
data.betas[data.betas>max.beta] <- max.beta
mcols(data.ranges) <- data.betas

# data.ranges and sample.ids objects are now ready for AMR search using ramr

Simulating data

ramr provides methods to create sets of random AMRs and to generate biologically relevant methylation beta values using real data sets as templates. The following code provides an example, however it is recommended to use a real experimental data (e.g. GSE51032) to create a test data set for assessing the performance of ramr or other AMR/DMR search engines. The results of parallel data generation are fully reproducible when the same seed has been set (thanks to doRNG::%dorng%).

# set the seed if reproducible results required
  set.seed(999)

# unique random AMRs
  amrs.unique <-
    simulateAMR(ramr.data, nsamples=25, regions.per.sample=2,
                min.cpgs=5, merge.window=1000, dbeta=0.2)

# non-unique AMRs outside of regions with unique AMRs
  amrs.nonunique <-
    simulateAMR(ramr.data, nsamples=4, exclude.ranges=amrs.unique,
                samples.per.region=2, min.cpgs=5, merge.window=1000)
  
# random noise outside of AMR regions
  noise <-
    simulateAMR(ramr.data, nsamples=25, regions.per.sample=20,
                exclude.ranges=c(amrs.unique, amrs.nonunique),
                min.cpgs=1, max.cpgs=1, merge.window=1, dbeta=0.5)

# "smooth" methylation data without AMRs (negative control)
  smooth.data <-
    simulateData(ramr.data, nsamples=25, cores=2)
#> Simulating data [3.119s]
  
# methylation data with AMRs and noise
  noisy.data <-
    simulateData(ramr.data, nsamples=25,
                 amr.ranges=c(amrs.unique, amrs.nonunique, noise), cores=2)
#> Simulating data [3.097s]
#> Introducing epimutations [0.105s]
  
# that's how regions look like
  library(gridExtra)
#> 
#> Attaching package: 'gridExtra'
#> 
#> The following object is masked from 'package:BiocGenerics':
#> 
#>     combine

  do.call("grid.arrange", c(plotAMR(noisy.data, amr.ranges=amrs.unique[1:4]), ncol=2))

  do.call("grid.arrange", c(plotAMR(noisy.data, amr.ranges=sort(amrs.nonunique)[1:8]), ncol=2))

  do.call("grid.arrange", c(plotAMR(noisy.data, amr.ranges=noise[1:4]), ncol=2))
#> geom_path: Each group consists of only one observation. Do you need to adjust the group
#> aesthetic?

  
# can we find them?
  system.time(found <- getAMR(noisy.data, ramr.method="beta", min.cpgs=5,
                              merge.window=1000, qval.cutoff=1e-2, cores=2))
#> Identifying AMRs [4.695s]
#>    user  system elapsed 
#>   7.078   1.183   4.866
  
# all possible regions
  all.ranges <- getUniverse(noisy.data, min.cpgs=5, merge.window=1000)
  
# true positives
  tp <- sum(found %over% c(amrs.unique, amrs.nonunique))
  
# false positives
  fp <- sum(found %outside% c(amrs.unique, amrs.nonunique))

# true negatives
  tn <- length(all.ranges %outside% c(amrs.unique, amrs.nonunique))
  
# false negatives
  fn <- sum(c(amrs.unique, amrs.nonunique) %outside% found)
  
# accuracy, MCC
  acc <- (tp+tn) / (tp+tn+fp+fn)
  mcc <- (tp*tn - fp*fn) / (sqrt(tp+fp)*sqrt(tp+fn)*sqrt(tn+fp)*sqrt(tn+fn))
  setNames(c(tp, fp, tn, fn), c("TP", "FP", "TN", "FN"))
#>  TP  FP  TN  FN 
#>  57   0 206   1
  setNames(c(acc, mcc), c("accuracy", "MCC"))
#>  accuracy       MCC 
#> 0.9962121 0.9889444

AMR identification

This code shows how to do basic analysis with ramr using provided data files:

# identify AMRs
amrs <- getAMR(ramr.data, ramr.samples, ramr.method="beta", min.cpgs=5,
               merge.window=1000, qval.cutoff=1e-3, cores=2)
#> Identifying AMRs [14.840s]

# inspect
sort(amrs)
#> GRanges object with 22 ranges and 5 metadata columns:
#>        seqnames          ranges strand |             revmap      ncpg      sample     dbeta
#>           <Rle>       <IRanges>  <Rle> |             <list> <integer> <character> <numeric>
#>    [1]     chr1   566172-569687      * |       17,18,19,...        15    sample95  0.498337
#>    [2]     chr1   874697-877876      * |    165,166,167,...        13    sample44  0.451475
#>    [3]     chr1   874697-877876      * |    165,166,167,...        13    sample45  0.457115
#>    [4]     chr1   874697-877876      * |    165,166,167,...        13    sample46  0.458498
#>    [5]     chr1 1095607-1106175      * |    620,621,622,...        49    sample66 -0.503085
#>    ...      ...             ...    ... .                ...       ...         ...       ...
#>   [18]     chr1 2200890-2203648      * | 2263,2264,2265,...        10    sample58 -0.505059
#>   [19]     chr1 2200890-2203648      * | 2263,2264,2265,...        10    sample59 -0.498849
#>   [20]     chr1 2200890-2203648      * | 2263,2264,2265,...        10    sample60 -0.500389
#>   [21]     chr1 2269871-2271665      * | 2410,2411,2412,...        10    sample71 -0.496600
#>   [22]     chr1 2443577-2453006      * | 2722,2723,2724,...        30    sample25 -0.484617
#>                pval
#>           <numeric>
#>    [1] 4.60616e-171
#>    [2]  1.30458e-07
#>    [3]  1.04111e-07
#>    [4]  8.03068e-08
#>    [5]  1.92980e-17
#>    ...          ...
#>   [18]  2.98503e-08
#>   [19]  4.05494e-08
#>   [20]  3.68870e-08
#>   [21]  1.54741e-17
#>   [22]  1.26772e-19
#>   -------
#>   seqinfo: 24 sequences from hg19 genome; no seqlengths

do.call("grid.arrange", c(plotAMR(ramr.data, ramr.samples, amrs[1:10]), ncol=2))

Again, the results of parallel processing are fully reproducible if the same seed has been set.

AMR annotation and enrichment analysis

If necessary, AMRs can be annotated to known genomic elements using R library annotatr 1 or tested for potential enrichment in epigenetic or other marks using R library LOLA 2

# annotating AMRs using R library annotatr
library(annotatr)
annotation.types <- c("hg19_cpg_inter", "hg19_cpg_islands", "hg19_cpg_shores",
                      "hg19_cpg_shelves", "hg19_genes_intergenic", "hg19_genes_promoters",
                      "hg19_genes_5UTRs", "hg19_genes_firstexons", "hg19_genes_3UTRs")
annotations <- build_annotations(genome='hg19', annotations=annotation.types)
#> Loading required package: GenomicFeatures
#> Loading required package: AnnotationDbi
#> Loading required package: Biobase
#> Welcome to Bioconductor
#> 
#>     Vignettes contain introductory material; view with 'browseVignettes()'. To cite
#>     Bioconductor, see 'citation("Biobase")', and for packages 'citation("pkgname")'.
#> 
#> 'select()' returned 1:1 mapping between keys and columns
#> Building promoters...
#> Building 1to5kb upstream of TSS...
#> Building intergenic...
#> Building 5UTRs...
#> Building 3UTRs...
#> Building exons...
#> Building first exons...
#> Building introns...
#> snapshotDate(): 2021-10-20
#> Building CpG islands...
#> loading from cache
#> Building CpG shores...
#> Building CpG shelves...
#> Building inter-CpG-islands...
amrs.annots <- annotate_regions(regions=amrs, annotations=annotations, ignore.strand=TRUE, quiet=FALSE)
#> Annotating...
summarize_annotations(annotated_regions=amrs.annots, quiet=FALSE)
#> Counting annotation types
#> # A tibble: 9 × 2
#>   annot.type                n
#>   <chr>                 <int>
#> 1 hg19_cpg_inter            4
#> 2 hg19_cpg_islands         10
#> 3 hg19_cpg_shelves          4
#> 4 hg19_cpg_shores          10
#> 5 hg19_genes_3UTRs          2
#> 6 hg19_genes_5UTRs          5
#> 7 hg19_genes_firstexons     8
#> 8 hg19_genes_intergenic     1
#> 9 hg19_genes_promoters      8
# generate the set of all possible genomic regions using sample data set and
# the same parameters as for AMR search
universe <- getUniverse(ramr.data, min.cpgs=5, merge.window=1000)

# enrichment analysis of AMRs using R library LOLA
library(LOLA)
# prepare the core database as described in vignettes
vignette("usingLOLACore")
# load the core database and perform the enrichment analysis
hg19.coredb <- loadRegionDB(system.file("LOLACore", "hg19", package="LOLA"))
runLOLA(amrs, universe, hg19.coredb, cores=1, redefineUserSets=TRUE)

Citing the ramr package

Oleksii Nikolaienko, Per Eystein Lønning, Stian Knappskog, ramr: an R/Bioconductor package for detection of rare aberrantly methylated regions, Bioinformatics, 2021;, btab586, https://doi.org/10.1093/bioinformatics/btab586

The data underlying ramr manuscript

Replication Data for: "ramr: an R package for detection of rare aberrantly methylated regions, https://doi.org/10.18710/ED8HSD

Session Info

sessionInfo()
#> R version 4.1.1 (2021-08-10)
#> Platform: x86_64-pc-linux-gnu (64-bit)
#> Running under: Ubuntu 20.04.3 LTS
#> 
#> Matrix products: default
#> BLAS:   /home/biocbuild/bbs-3.14-bioc/R/lib/libRblas.so
#> LAPACK: /home/biocbuild/bbs-3.14-bioc/R/lib/libRlapack.so
#> 
#> locale:
#>  [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C               LC_TIME=en_GB             
#>  [4] LC_COLLATE=C               LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8   
#>  [7] LC_PAPER=en_US.UTF-8       LC_NAME=C                  LC_ADDRESS=C              
#> [10] LC_TELEPHONE=C             LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C       
#> 
#> attached base packages:
#> [1] parallel  stats4    stats     graphics  grDevices utils     datasets  methods   base     
#> 
#> other attached packages:
#>  [1] org.Hs.eg.db_3.14.0                     TxDb.Hsapiens.UCSC.hg19.knownGene_3.2.2
#>  [3] GenomicFeatures_1.46.0                  AnnotationDbi_1.56.0                   
#>  [5] Biobase_2.54.0                          annotatr_1.20.0                        
#>  [7] gridExtra_2.3                           ramr_1.2.0                             
#>  [9] doRNG_1.8.2                             rngtools_1.5.2                         
#> [11] doParallel_1.0.16                       iterators_1.0.13                       
#> [13] foreach_1.5.1                           GenomicRanges_1.46.0                   
#> [15] GenomeInfoDb_1.30.0                     IRanges_2.28.0                         
#> [17] S4Vectors_0.32.0                        BiocGenerics_0.40.0                    
#> 
#> loaded via a namespace (and not attached):
#>   [1] colorspace_2.0-2              rjson_0.2.20                  ellipsis_0.3.2               
#>   [4] EnvStats_2.4.0                XVector_0.34.0                farver_2.1.0                 
#>   [7] bit64_4.0.5                   interactiveDisplayBase_1.32.0 fansi_0.5.0                  
#>  [10] xml2_1.3.2                    codetools_0.2-18              cachem_1.0.6                 
#>  [13] knitr_1.36                    jsonlite_1.7.2                Rsamtools_2.10.0             
#>  [16] dbplyr_2.1.1                  png_0.1-7                     shiny_1.7.1                  
#>  [19] BiocManager_1.30.16           readr_2.0.2                   compiler_4.1.1               
#>  [22] httr_1.4.2                    assertthat_0.2.1              Matrix_1.3-4                 
#>  [25] fastmap_1.1.0                 cli_3.0.1                     later_1.3.0                  
#>  [28] htmltools_0.5.2               prettyunits_1.1.1             tools_4.1.1                  
#>  [31] gtable_0.3.0                  glue_1.4.2                    GenomeInfoDbData_1.2.7       
#>  [34] reshape2_1.4.4                dplyr_1.0.7                   rappdirs_0.3.3               
#>  [37] Rcpp_1.0.7                    jquerylib_0.1.4               vctrs_0.3.8                  
#>  [40] Biostrings_2.62.0             rtracklayer_1.54.0            xfun_0.27                    
#>  [43] stringr_1.4.0                 mime_0.12                     lifecycle_1.0.1              
#>  [46] restfulr_0.0.13               XML_3.99-0.8                  AnnotationHub_3.2.0          
#>  [49] zlibbioc_1.40.0               scales_1.1.1                  BSgenome_1.62.0              
#>  [52] hms_1.1.1                     promises_1.2.0.1              MatrixGenerics_1.6.0         
#>  [55] SummarizedExperiment_1.24.0   yaml_2.2.1                    curl_4.3.2                   
#>  [58] memoise_2.0.0                 ggplot2_3.3.5                 sass_0.4.0                   
#>  [61] biomaRt_2.50.0                stringi_1.7.5                 RSQLite_2.2.8                
#>  [64] BiocVersion_3.14.0            highr_0.9                     BiocIO_1.4.0                 
#>  [67] filelock_1.0.2                optimx_2021-10.12             BiocParallel_1.28.0          
#>  [70] rlang_0.4.12                  pkgconfig_2.0.3               matrixStats_0.61.0           
#>  [73] bitops_1.0-7                  evaluate_0.14                 lattice_0.20-45              
#>  [76] purrr_0.3.4                   GenomicAlignments_1.30.0      labeling_0.4.2               
#>  [79] bit_4.0.4                     tidyselect_1.1.1              plyr_1.8.6                   
#>  [82] magrittr_2.0.1                R6_2.5.1                      generics_0.1.1               
#>  [85] DelayedArray_0.20.0           DBI_1.1.1                     withr_2.4.2                  
#>  [88] pillar_1.6.4                  KEGGREST_1.34.0               RCurl_1.98-1.5               
#>  [91] tibble_3.1.5                  crayon_1.4.1                  utf8_1.2.2                   
#>  [94] BiocFileCache_2.2.0           tzdb_0.1.2                    rmarkdown_2.11               
#>  [97] progress_1.2.2                ExtDist_0.6-4                 grid_4.1.1                   
#> [100] blob_1.2.2                    digest_0.6.28                 xtable_1.8-4                 
#> [103] regioneR_1.26.0               httpuv_1.6.3                  numDeriv_2016.8-1.1          
#> [106] munsell_0.5.0                 bslib_0.3.1

References

  1. Raymond G Cavalcante, Maureen A Sartor, annotatr: genomic regions in context, Bioinformatics, Volume 33, Issue 15, 01 August 2017, Pages 2381–2383, https://doi.org/10.1093/bioinformatics/btx183↩

  2. Nathan C. Sheffield, Christoph Bock, LOLA: enrichment analysis for genomic region sets and regulatory elements in R and Bioconductor, Bioinformatics, Volume 32, Issue 4, 15 February 2016, Pages 587–589, https://doi.org/10.1093/bioinformatics/btv612↩