Contents

Package: metagene
Modified: 18 september, 2015
Compiled: Mon Oct 30 20:01:49 2017
License: Artistic-2.0 | file LICENSE

1 Introduction

This package produces metagene-like plots to compare the behavior of DNA-interacting proteins at selected groups of features. A typical analysis can be done in viscinity of transcription start sites (TSS) of genes or at any regions of interest (such as enhancers). Multiple combinations of group of features and/or group of bam files can be compared in a single analysis. Bootstraping analysis is used to compare the groups and locate regions with statistically different enrichment profiles. In order to increase the sensitivity of the analysis, alignment data is used instead of peaks produced with peak callers (i.e.: MACS2 or PICS). The metagene package uses bootstrap to obtain a better estimation of the mean enrichment and the confidence interval for every group of samples.

This vignette will introduce all the main features of the metagene package.

2 Loading the metagene package

library(metagene)

3 Inputs

3.1 Alignment files (BAM files)

There is no hard limit in the number of BAM files that can be included in an analysis (but with too many BAM files, memory may become an issue). BAM files must be indexed. For instance, if you use a file names file.bam, a file named file.bam.bai or file.baimust be present in the same directory.

The path (relative or absolute) to the BAM files must be in a vector:

bam_files <- get_demo_bam_files()
bam_files
## [1] "/tmp/RtmpIzf1Dz/Rinst786712897901/metagene/extdata/align1_rep1.bam"
## [2] "/tmp/RtmpIzf1Dz/Rinst786712897901/metagene/extdata/align1_rep2.bam"
## [3] "/tmp/RtmpIzf1Dz/Rinst786712897901/metagene/extdata/align2_rep1.bam"
## [4] "/tmp/RtmpIzf1Dz/Rinst786712897901/metagene/extdata/align2_rep2.bam"
## [5] "/tmp/RtmpIzf1Dz/Rinst786712897901/metagene/extdata/ctrl.bam"

For this demo, we have 2 samples (each with 2 replicates). It is also possible to use a named vector to add your own names to each BAM files:

named_bam_files <- bam_files
names(named_bam_files) <- letters[seq_along(bam_files)]
named_bam_files
##                                                                    a 
## "/tmp/RtmpIzf1Dz/Rinst786712897901/metagene/extdata/align1_rep1.bam" 
##                                                                    b 
## "/tmp/RtmpIzf1Dz/Rinst786712897901/metagene/extdata/align1_rep2.bam" 
##                                                                    c 
## "/tmp/RtmpIzf1Dz/Rinst786712897901/metagene/extdata/align2_rep1.bam" 
##                                                                    d 
## "/tmp/RtmpIzf1Dz/Rinst786712897901/metagene/extdata/align2_rep2.bam" 
##                                                                    e 
##        "/tmp/RtmpIzf1Dz/Rinst786712897901/metagene/extdata/ctrl.bam"

Using named BAM files can simplify the use of the metagene helper functions and the creation of the design.

3.2 Genomic regions

3.2.1 BED files

To compare custom regions of interest, it is possible to use a list of one or more BED files.

regions <- get_demo_regions()
regions
## [1] "/tmp/RtmpIzf1Dz/Rinst786712897901/metagene/extdata/list1.bed"
## [2] "/tmp/RtmpIzf1Dz/Rinst786712897901/metagene/extdata/list2.bed"

The name of the files (without the extension) will be used to name each groups.

metagene also support the narrowPeak and the broadPeak.

3.2.2 GRanges or GRangesList objects - Regions

As an alternative to a list of BED files, GRanges or GRangesList objects can be used.

3.2.3 Available datasets

Some common datasets are already available with the metagene package:

  • promoters_hg19
  • promoters_hg18
  • promoters_mm10
  • promoters_mm9
data(promoters_hg19)
promoters_hg19
## GRanges object with 23056 ranges and 1 metadata column:
##         seqnames                 ranges strand |     gene_id
##            <Rle>              <IRanges>  <Rle> | <character>
##       1    chr19 [ 58873215,  58875214]      - |           1
##      10     chr8 [ 18247755,  18249754]      + |          10
##     100    chr20 [ 43279377,  43281376]      - |         100
##    1000    chr18 [ 25756446,  25758445]      - |        1000
##   10000     chr1 [244005887, 244007886]      - |       10000
##     ...      ...                    ...    ... .         ...
##    9991     chr9 [115094945, 115096944]      - |        9991
##    9992    chr21 [ 35735323,  35737322]      + |        9992
##    9993    chr22 [ 19108968,  19110967]      - |        9993
##    9994     chr6 [ 90538619,  90540618]      + |        9994
##    9997    chr22 [ 50963906,  50965905]      - |        9997
##   -------
##   seqinfo: 93 sequences (1 circular) from hg19 genome

For more details about each datasets, please refer to their documentation (i.e.:?promoters_hg19).

3.3 Design groups

A design group contains a set of BAM files that, when put together, represent a logical analysis. Furthermore, a design group contains the relationship between every BAM files present. Samples (with or without replicates) and controls can be assigned to a same design group. There can be as many groups as necessary. A BAM file can be assigned to more than one group.

To represent the relationship between every BAM files, design groups must have the following columns:

  • The list of paths to every BAM files related to an analysis.
  • One column per group of files (replicates and/or controls).

There is two possible way to create design groups, by reading a file or by directly creating a design object in R.

3.3.1 Design groups from a file

Design groups can be loaded into the metagene package by using a text file. As the relationship between BAM files as to be specified, the following columns are mandatory:

  • First column: The list of paths (absolute or relative) to every BAM files for all the design groups, the BAM filenames or the BAM names (if a named BAM. file was used).
  • Following columns: One column per design group (replicates and/or controls). The column can take only three values:
    • 0: ignore file
    • 1: input
    • 2: control

The file must also contain a header. It is recommanded to use Samples for the name of the first column, but the value is not checked. The other columns in the design file will be used for naming design groups, and must be unique.

fileDesign <- system.file("extdata/design.txt", package="metagene")
design <- read.table(fileDesign, header=TRUE, stringsAsFactors=FALSE)
design$Samples <- paste(system.file("extdata", package="metagene"),
                        design$Samples, sep="/")
kable(design)
Samples align1 align2
/tmp/RtmpIzf1Dz/Rinst786712897901/metagene/extdata/align1_rep1.bam 1 0
/tmp/RtmpIzf1Dz/Rinst786712897901/metagene/extdata/align1_rep2.bam 1 0
/tmp/RtmpIzf1Dz/Rinst786712897901/metagene/extdata/align2_rep1.bam 0 1
/tmp/RtmpIzf1Dz/Rinst786712897901/metagene/extdata/align2_rep2.bam 0 1
/tmp/RtmpIzf1Dz/Rinst786712897901/metagene/extdata/ctrl.bam 2 2

3.3.2 Design groups from R

It is not obligatory to use a design file, you can create the design data.frame using your prefered method (as long as the restrictions on the values mentioned previously are respected).

For instance, the previous design data.frame could have been create directly in R:

design <- data.frame(Samples = c("align1_rep1.bam", "align1_rep2.bam",
                    "align2_rep1.bam", "align2_rep2.bam", "ctrl.bam"),
                    align1 = c(1,1,0,0,2), align2 = c(0,0,1,1,2))
design$Samples <- paste0(system.file("extdata", package="metagene"), "/",
                        design$Samples)
kable(design)
Samples align1 align2
/tmp/RtmpIzf1Dz/Rinst786712897901/metagene/extdata/align1_rep1.bam 1 0
/tmp/RtmpIzf1Dz/Rinst786712897901/metagene/extdata/align1_rep2.bam 1 0
/tmp/RtmpIzf1Dz/Rinst786712897901/metagene/extdata/align2_rep1.bam 0 1
/tmp/RtmpIzf1Dz/Rinst786712897901/metagene/extdata/align2_rep2.bam 0 1
/tmp/RtmpIzf1Dz/Rinst786712897901/metagene/extdata/ctrl.bam 2 2

4 Analysis steps

A typical metagene analysis will consist steps:

4.1 Minimal analysis

A minimal metagene analysis can be performed in 2 steps:

  1. Initialization (the new function).
  2. plot
regions <- get_demo_regions()
bam_files <- get_demo_bam_files()
# Initialization
mg <- metagene$new(regions = regions, bam_files = bam_files)
# Plotting
mg$plot(title = "Demo metagene plot")
## produce data table : ChIP-Seq
## produce data frame : ChIP-Seq
## Plot : ChIP-Seq

As you can see, it is not mandatory to explicitly call each step of the metagene analysis. For instance, in the previous example, the plot function call the other steps automatically with default values (the next section will describe the steps in more details).

In this specific case, the plot is messy since by default metagene will produce a curve for each possible combinations of BAM file and regions. Since we have 5 BAM files and 2 regions, this gives us 10 curves.

If we want more control on how every step of the analysis are performed, we have to call each functions directly.

4.2 Complete analysis

In order to fully control every step of a metagene analysis, it is important to understand how a complete analysis is performed. If we are satisfied with the default values, it is not mandatory to explicitly call every step (as was shown in the previous section).

4.2.1 Initialization

During this step, the coverages for every regions specified are extracted from every BAM files. More specifically, a new GRanges is created by combining all the regions specified with the regions param of the new function.

regions <- get_demo_regions()
bam_files <- get_demo_bam_files()
mg <- metagene$new(regions = regions, bam_files = bam_files)

4.2.2 Producing the table

To produce the table, coverages (produced from Genomics regions (.BED), Alignment Files (.BAM) and Design Sheet) was treated for noise removal and normalized. Furthermore, to reduce the computation time during the following steps, the positions are also binned. Regions, designs, bins, associated values and orientation of strands are pulled into a data.table called ‘table’ and accessible thanks to the getter get_table.

We can control the size of the bins with the bin_count argument. By default, a bin_count of 100 will be used during this step.

mg$produce_table()
## produce data table : ChIP-Seq

We can also use the design we produced earlier to remove background signal and combine replicates:

mg$produce_table(design = design)
## produce data table : ChIP-Seq

4.2.3 Producing the data.frame

The metagene plot are produced using the ggplot2 package, which require a data.frame as input. During this step, the values of the ribbon are calculated. Metagene uses “bootstrap” to obtain a better estimation of the mean of enrichment for every positions in each groups.

mg$produce_data_frame()
## produce data frame : ChIP-Seq

4.2.4 Plotting

During this step, metagene will use the data.frame to plot the calculated values using ggplot2. We show a subset of the regions by using the region_names and design_names parameter. The region_names correspond to the names of the regions used during the initialization. The design_name will vary depending if a design was added. If no design was added, this param correspond to the BAM name or BAM filenames. Otherwise, we have to use the names of the columns from the design.

mg$plot(region_names = "list1", title = "Demo plot subset")
## Plot : ChIP-Seq