1 Licensing and citing

The RJMCMCNucleosomes package and the underlying RJMCMCNucleosomes code are distributed under the Artistic license 2.0. You are free to use and redistribute this software.

If you use this package for a publication, we would ask you to cite the following:

Samb R, Khadraoui K, Belleau P, et al. (2015) Using informative Multinomial-Dirichlet prior in a t-mixture with reversible jump estimation of nucleosome positions for genome-wide profiling. Statistical Applications in Genetics and Molecular Biology. Volume 14, Issue 6, Pages 517-532, ISSN (Online) 1544-6115, ISSN (Print) 2194-6302, December 2015, doi:10.1515/sagmb-2014-0098

2 Introduction

Global gene expression patterns are established and maintained by the concerted actions of transcription factors (TFs) and the proteins that constitute chromatin. The key structural element of chromatin is the nucleosome, which consists of an octameric histone core wrapped by 146 bps of DNA and connected to its neighbour by approximately 10-80 pbs of linker DNA (Polishko et al. 2012).

The literature on nucleosome positioning commonly focuses on frequentist inferences within parametric approaches (see for instance Chen et al. (2010) and Xi et al. (2010)). In those works, the detection of nucleosome positions is done using a hidden Markov model with an assumed known order.

The RJMCMCNucleosomes package is an implementation of a fully Bayesian hierarchical model for profiling of nucleosome positions based on high-throughput short-read data (MNase-Seq data). The implementation is based on a strategy which incorporates four aspects. First, it jointly models local concentrations of directional reads. Second, it uses a Multinomial-Dirichlet model in the construction of an informative prior distribution coupled to a t-mixture model with unknown degrees of freedom. Third, the number of nucleosomes is considered to be a random variable and refers to a prior distribution. Fourth, the unknown parameters are simultaneously using the reversible jump Markov chain Monte Carlo (RJMCMC) simulation technique (see for instance Green (1995) and Richardson and Green (1997)).

Detailed information about the model can be found in the article mentioned in the citing section.

3 Loading the RJMCMC package

As with any R package, the RJMCMCNucleosomes package should first be loaded with the following command:

library(RJMCMCNucleosomes)

4 RJMCMCNucleosomes analysis

A typical RJMCMCNucleosomes analysis consists of the following steps:

1. Segment the analysed region into candidate regions that have sufficient aligned reads. The initial region cannot be wider than one chromosome.
2. Estimate nucleosome positions for each region.
3. Regroup all regions together. The final region cannot be wider than one chromosome.
4. Post-process predictions of the regrouped region to revise certain predictions.

5 RJMCMCNucleosomes analysis step by step

A synthetic nucleosome sample containing 100 nucleosomes (80 well-positioned + 20 fuzzy) has been created using the Bioconductor package nucleoSim. This synthetic sample will be used throughout this analysis.

## Load nucleoSim package
library(nucleoSim)

val.num       <- 50     ### Number of well-positioned nucleosomes
val.del       <- 10     ### Number of well-positioned nucleosomes to delete
val.var       <- 30     ### variance associated to well-positioned nucleosomes
val.fuz       <- 10     ### Number of fuzzy nucleosomes
val.fuz.var   <- 50     ### variance associated to fuzzy nucleosomes
val.max.cover <- 70     ### Maximum coverage for one nucleosome
val.nuc.len   <- 147    ### Distance between nucleosomes
val.len.var   <- 10     ### Variance associated to the length of the reads
val.lin.len   <- 20     ### The length of the DNA linker
val.rnd.seed  <- 100    ### Set seed when result needs to be reproducible
val.offset    <- 10000  ### The number of bases used to offset 
                        ### all nucleosomes and reads

## Create sample using a Normal distribution
sample <- nucleoSim::syntheticNucReadsFromDist(wp.num=val.num,
                                    wp.del=val.del, wp.var=val.var,
                                    fuz.num=val.del, fuz.var=val.fuz.var,
                                    max.cover=val.max.cover, 
                                    nuc.len=val.nuc.len,
                                    len.var=val.len.var, 
                                    lin.len=val.lin.len,
                                    rnd.seed=val.rnd.seed,
                                    distr="Normal", offset=val.offset)

## Create visual representation of the synthetic nucleosome sample
plot(sample)

## Load needed packages
library(GenomicRanges)

## Transform sample dataset into GRanges object
sampleGRanges <- GRanges(seqnames = sample$dataIP$chr, 
                        ranges = IRanges(start = sample$dataIP$start, 
                                        end = sample$dataIP$end), 
                        strand = sample$dataIP$strand)

## Segment sample into candidate regions
sampleSegmented <- segmentation(reads = sampleGRanges, zeta = 147, 
                                delta = 40, maxLength = 1000)

## Number of segments created
length(sampleSegmented)
## [1] 11

## Extract the first segment 
segment01 <- sampleSegmented[[1]]

## Run RJMCMC analysis
## A higher number of iterations is recommanded for real analysis
resultSegment01 <- rjmcmc(reads  = segment01, nbrIterations = 100000, 
                            lambda = 3, kMax = 30,
                            minInterval = 100, maxInterval = 200, 
                            minReads = 5, vSeed = 1921)

## Print the predicted nucleosomes for the first segment
resultSegment01
## RJMCMCNucleosomes - Predicted nucleosomes
## 
## Call:
## rjmcmc(reads = segment01, nbrIterations = 1e+05, kMax = 30, lambda = 3, 
##     minInterval = 100, maxInterval = 200, minReads = 5, vSeed = 1921)
## 
## Number of nucleosomes:
## [1] 5
## 
## Nucleosomes positions:
## GRanges object with 5 ranges and 0 metadata columns:
##            seqnames    ranges strand
##               <Rle> <IRanges>  <Rle>
##   [1] chr_SYNTHETIC     10075      *
##   [2] chr_SYNTHETIC     10245      *
##   [3] chr_SYNTHETIC     10409      *
##   [4] chr_SYNTHETIC     10569      *
##   [5] chr_SYNTHETIC     10740      *
##   -------
##   seqinfo: 1 sequence from an unspecified genome; no seqlengths

## The directory containing the results of all segments
## On RDS file has been created for each segment
directory <- system.file("extdata", "demo_vignette", 
                            package = "RJMCMCNucleosomes")

## Merging predicted nucleosomes from all segments
resultsAllMerged <- mergeAllRDSFilesFromDirectory(directory)

resultsAllMerged
## 
## RJMCMCNucleosomes - Predicted nucleosomes
## 
## Number of nucleosomes:
## [1] 63
## 
## Nucleosomes positions:
## GRanges object with 63 ranges and 0 metadata columns:
##             seqnames    ranges strand
##                <Rle> <IRanges>  <Rle>
##    [1] chr_SYNTHETIC     10075      *
##    [2] chr_SYNTHETIC     10245      *
##    [3] chr_SYNTHETIC     10407      *
##    [4] chr_SYNTHETIC     10571      *
##    [5] chr_SYNTHETIC     10742      *
##    ...           ...       ...    ...
##   [59] chr_SYNTHETIC     17421      *
##   [60] chr_SYNTHETIC     17589      *
##   [61] chr_SYNTHETIC     17756      *
##   [62] chr_SYNTHETIC     17771      *
##   [63] chr_SYNTHETIC     18257      *
##   -------
##   seqinfo: 1 sequence from an unspecified genome; no seqlengths

## Split reads from the initial sample data into forward and reverse subsets
reads <- GRanges(sample$dataIP)

## Number of nucleosomes before the post-treatment
resultsAllMerged$k
## [1] 63

## Use the post-treatment function
resultsPostTreatment <- postTreatment(reads = reads,
                            resultRJMCMC = resultsAllMerged,
                            extendingSize = 15,
                            chrLength = max(start(reads), end(reads)) + 1000)


## Number of nucleosomes after the post-treatment
length(resultsPostTreatment)
## [1] 51

## Extract reads to create a GRanges
reads <-GRanges(sample$dataIP)

## Merge predictions from before post-treatment and after post-treatment in 
## a list so that both results will be shown in graph
# resultsBeforeAndAfter <- list(Sample = c(sample$wp$nucleopos,
#                                             sample$fuz$nucleopos),
#                                BeforePostTreatment = resultsAllMerged$mu,
#                                AfterPostTreatment = resultsPostTreatment)
resultsBeforeAndAfter <- GRangesList(Sample = GRanges(
        rep("chr_SYNTHETIC", 
            length(c(sample$wp$nucleopos,sample$fuz$nucleopos))), 
        ranges = IRanges(start=c(sample$wp$nucleopos,sample$fuz$nucleopos),
                        end=c(sample$wp$nucleopos,sample$fuz$nucleopos)), 
        strand=rep("*", length(c(sample$wp$nucleopos,
                                sample$fuz$nucleopos)))),
                        BeforePostTreatment = resultsAllMerged$mu,
                        AfterPostTreatment = resultsPostTreatment)
## Create graph using nucleosome positions and reads
## The plot will shows : 
## 1. nucleosomes from the sample
## 2. nucleosomes detected by rjmcmc() function
## 3. nucleosomes obtained after post-treament 
plotNucleosomes(nucleosomePositions = resultsBeforeAndAfter, reads = reads, 
                    names=c("Sample", "RJMCMC", "After Post-Treatment"))

6 RJMCMCNucleosomes analysis of one chromosome in one step

The rjmcmcCHR can analyse an entire chromosome by automatically split the reads into segments, run rjmcmc on each segment, merge and post-process the results. The intermediary steps are conserved in a directory set by the dirOut parameter.

On real chromosome data, the rjmcmcCHR can take some time to execute. We strongly suggest running it on a multi-core computer and to use a maximum of cores available by setting the nbCores parameter.

## Load synthetic dataset of reads
data(syntheticNucleosomeReads)

## Number of reads in the dataset
nrow(syntheticNucleosomeReads$dataIP)

## Use the dataset to create a GRanges object
sampleGRanges <- GRanges(syntheticNucleosomeReads$dataIP)

## All reads are related to one chromosome called "chr_SYNTHETIC"
seqnames(sampleGRanges)

## Run RJMCMC on all reads
result <- rjmcmcCHR(reads = sampleGRanges, 
                seqName = "chr_SYNTHETIC", dirOut = "testRJMCMCCHR",
                zeta = 147, delta=50, maxLength=1200,
                nbrIterations = 500, lambda = 3, kMax = 30,
                minInterval = 146, maxInterval = 292, minReads = 5,
                vSeed = 10113, nbCores = 2, saveAsRDS = FALSE, 
                saveSEG = FALSE)

result

When the saveSEG parameter is set to TRUE, the segments created during the segmentation step are saved in RDS files. To save the results for each segment, the saveRDS parameter has to be set to TRUE.

7 Session info

Here is the output of sessionInfo() on the system on which this document was compiled:

## R version 3.5.0 (2018-04-23)
## Platform: x86_64-pc-linux-gnu (64-bit)
## Running under: Ubuntu 16.04.4 LTS
## 
## Matrix products: default
## BLAS: /home/biocbuild/bbs-3.7-bioc/R/lib/libRblas.so
## LAPACK: /home/biocbuild/bbs-3.7-bioc/R/lib/libRlapack.so
## 
## locale:
##  [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C              
##  [3] LC_TIME=en_US.UTF-8        LC_COLLATE=C              
##  [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8   
##  [7] LC_PAPER=en_US.UTF-8       LC_NAME=C                 
##  [9] LC_ADDRESS=C               LC_TELEPHONE=C            
## [11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C       
## 
## attached base packages:
## [1] stats4    parallel  stats     graphics  grDevices utils     datasets 
## [8] methods   base     
## 
## other attached packages:
## [1] nucleoSim_1.8.0         RJMCMCNucleosomes_1.4.0 GenomicRanges_1.32.0   
## [4] GenomeInfoDb_1.16.0     IRanges_2.14.0          S4Vectors_0.18.0       
## [7] BiocGenerics_0.26.0     BiocStyle_2.8.0        
## 
## loaded via a namespace (and not attached):
##  [1] Rcpp_0.12.16                compiler_3.5.0             
##  [3] XVector_0.20.0              consensusSeekeR_1.8.0      
##  [5] bitops_1.0-6                tools_3.5.0                
##  [7] zlibbioc_1.26.0             digest_0.6.15              
##  [9] evaluate_0.10.1             lattice_0.20-35            
## [11] Matrix_1.2-14               DelayedArray_0.6.0         
## [13] yaml_2.1.18                 xfun_0.1                   
## [15] GenomeInfoDbData_1.1.0      rtracklayer_1.40.0         
## [17] stringr_1.3.0               knitr_1.20                 
## [19] Biostrings_2.48.0           rprojroot_1.3-2            
## [21] grid_3.5.0                  Biobase_2.40.0             
## [23] XML_3.98-1.11               BiocParallel_1.14.0        
## [25] rmarkdown_1.9               bookdown_0.7               
## [27] magrittr_1.5                backports_1.1.2            
## [29] Rsamtools_1.32.0            htmltools_0.3.6            
## [31] matrixStats_0.53.1          GenomicAlignments_1.16.0   
## [33] SummarizedExperiment_1.10.0 stringi_1.1.7              
## [35] RCurl_1.95-4.10