In Bioconductor/RaggedExperiment: Representation of Sparse Experiments and Assays Across Samples

Introduction

The r Biocpkg("RaggedExperiment") package provides a flexible data representation for copy number, mutation and other ragged array schema for genomic location data. The output of Allele-Specific Copy number Analysis of Tumors (ASCAT) can be classed as a ragged array and contains whole genome allele-specific copy number information for each sample in the analysis. For more information on ASCAT and guidelines on how to generate ASCAT data please see the ASCAT website and github. To carry out further analysis of the ASCAT data, utilising the functionalities of RaggedExperiment, the ASCAT data must undergo a number of operations to get it in the correct format for use with RaggedExperiment.

Installation

if (!require("BiocManager"))
    install.packages("BiocManager")

BiocManager::install("RaggedExperiment")

Loading the package:

library(RaggedExperiment)
library(GenomicRanges)

Structure of ASCAT data

The data shown below is the output obtained from ASCAT. ASCAT takes Log R Ratio (LRR) and B Allele Frequency (BAF) files and derives the allele-specific copy number profiles of tumour cells, accounting for normal cell admixture and tumour aneuploidy. It should be noted that if working with raw CEL files, the first step is to preprocess the CEL files using the PennCNV-Affy pipeline described here. The PennCNV-Affy pipeline produces the LRR and BAF files used as inputs for ASCAT.

Depending on user preference, the output of ASCAT can be multiple files, each one containing allele-specific copy number information for one of the samples processed in an ASCAT run, or can be a single file containing allele-specific copy number information for all samples processed in an ASCAT run.

Let's load up and have a look at ASCAT data that contains copy number information for just one sample i.e. sample1. Here we load up the data, check that it only contains allele-specific copy number calls for 1 sample and look at the first 10 rows of the dataframe.

ASCAT_data_S1 <- read.delim(
    system.file(
        "extdata", "ASCAT_Sample1.txt",
        package = "RaggedExperiment", mustWork = TRUE
    ),
    header = TRUE
)

unique(ASCAT_data_S1$sample)

head(ASCAT_data_S1, n = 10)

Now let's load up and have a look at ASCAT data that contains copy number information for the three processed samples i.e. sample1, sample2 and sample3. Here we load up the data, check that it contains allele-specific copy number calls for the 3 samples and look at the first 10 rows of the dataframe. We also note that as expected the copy number calls for sample1 are the same as above.

ASCAT_data_All <- read.delim(
    system.file(
        "extdata", "ASCAT_All_Samples.txt",
        package = "RaggedExperiment", mustWork = TRUE
    ),
    header = TRUE
)

unique(ASCAT_data_All$sample)

head(ASCAT_data_All, n = 10)

From the output above we can see that the ASCAT data has 6 columns named sample, chr, startpos, endpos, nMajor and nMinor. These correspond to the sample ID, chromosome, the start position and end position of the genomic ranges and the copy number of the major and minor alleles i.e. the homologous chromosomes.

Converting ASCAT data to GRanges format

The RaggedExperiment class derives from a GRangesList representation and can take a GRanges object, a GRangesList or a list of Granges as inputs. To be able to use the ASCAT data in RaggedExperiment we must convert the ASCAT data into GRanges format. Ideally, we want each of our GRanges objects to correspond to an individual sample.

ASCAT to GRanges objects

In the case where the ASCAT data has only 1 sample it is relatively simple to produce a GRanges object.

sample1_ex1 <- GRanges(
    seqnames = Rle(paste0("chr", ASCAT_data_S1$chr)),
    ranges = IRanges(start = ASCAT_data_S1$startpos, end = ASCAT_data_S1$endpos),
    strand = Rle(strand("*")),
    nmajor = ASCAT_data_S1$nMajor,
    nminor = ASCAT_data_S1$nMinor
)

sample1_ex1

Here we create a GRanges object by taking each column of the ASCAT data and assigning them to the appropriate argument in the GRanges function. From above we can see that the chromosome information is prefixed with "chr" and becomes the seqnames column, the start and end positions are combined into an IRanges object and given to the ranges argument, the strand column contains a * for each entry as we don't have strand information and the metadata columns contain the allele-specific copy number calls and are called nmajor and nminor. The GRanges object we have just created contains 41 ranges (rows) and 2 metadata columns.

Another way that we can easily convert our ASCAT data, containing 1 sample, to a GRanges object is to use the makeGRangesFromDataFrame function from the GenomicsRanges package. Here we indicate what columns in our data correspond to the chromosome (given to the seqnames argument), start and end positions (start.field and end.field arguments), whether to ignore strand information and assign all entries * (ignore.strand) and also whether to keep the other columns in the dataframe, nmajor and nminor, as metadata columns (keep.extra.columns).

sample1_ex2 <- makeGRangesFromDataFrame(
    ASCAT_data_S1[,-c(1)],
    ignore.strand=TRUE,
    seqnames.field="chr",
    start.field="startpos",
    end.field="endpos",
    keep.extra.columns=TRUE
)

sample1_ex2

In the case where the ASCAT data contains more than 1 sample you can first use the split function to split the whole dataframe into multiple dataframes, one for each sample, and then create a GRanges object for each dataframe. Code to split the dataframe, based on sample ID, is given below and then the same procedure used to produce sample1_ex2 can be implemented to create the GRanges object. Alternatively, an easier and more efficient way to do this is to use the makeGRangesListFromDataFrame function from the GenomicsRanges package. This will be covered in the next section.

sample_list <- split(
    ASCAT_data_All,
    f = ASCAT_data_All$sample
)

ASCAT to GRangesList instance

To produce a GRangesList instance from the ASCAT dataframe we can use the makeGRangesListFromDataFrame function. This function takes the same arguments as the makeGRangesFromDataFrame function used above, but also has an argument specifying how the rows of the df are split (split.field). Here we will split on sample. This function can be used in cases where the ASCAT data contains only 1 sample or where it contains multiple samples.

Using makeGRangesListFromDataFrame to create a list of GRanges objects where ASCAT data has only 1 sample:

sample_list_GRanges_ex1 <- makeGRangesListFromDataFrame(
    ASCAT_data_S1,
    ignore.strand=TRUE,
    seqnames.field="chr",
    start.field="startpos",
    end.field="endpos",
    keep.extra.columns=TRUE,
    split.field = "sample"
)

sample_list_GRanges_ex1

Using makeGRangesListFromDataFrame to create a list of GRanges objects where ASCAT data has multiple samples:

sample_list_GRanges_ex2 <- makeGRangesListFromDataFrame(
    ASCAT_data_All,
    ignore.strand=TRUE,
    seqnames.field="chr",
    start.field="startpos",
    end.field="endpos",
    keep.extra.columns=TRUE,
    split.field = "sample"
)

sample_list_GRanges_ex2

Each GRanges object in the list can then be accessed using square bracket notation.

sample1_ex3 <- sample_list_GRanges_ex2[[1]]

sample1_ex3

Another way we can produce a GRangesList instance is to use the GRangesList function. This function creates a list that contains all our GRanges objects. This is straightforward in that we use the GRangesList function with our GRanges objects as named or unnamed inputs. Below we have created a list that includes 1 GRanges objects, created in section 4.1., corresponding to sample1.

sample_list_GRanges_ex3 <- GRangesList(
    sample1 = sample1_ex1
)

sample_list_GRanges_ex3

Constructing a RaggedExperiment object from ASCAT output

Now we have created the GRanges objects and GRangesList instances we can easily use RaggedExperiment.

Using GRanges objects

From above we have a GRanges object derived from the ASCAT data containing 1 sample i.e. sample1_ex1 / sample1_ex2 and the capabilities to produce individual GRanges objects derived from the ASCAT data containing 3 samples. We can now use these GRanges objects as inputs to RaggedExperiment. Note that we create column data colData to describe the samples.

Using GRanges object where ASCAT data only has 1 sample:

colDat_1 = DataFrame(id = 1)

ragexp_1 <- RaggedExperiment(
    sample1 = sample1_ex2,
    colData = colDat_1
)

ragexp_1

In the case where you have multiple GRanges objects, corresponding to different samples, the code is similar to above. Each sample is inputted into the RaggedExperiment function and colDat_1 corresponds to the id for each sample i.e. 1, 2 and 3, if 3 samples are provided.

Using a GRangesList instance

From before we have a GRangesList derived from the ASCAT data containing 1 sample i.e. sample_list_GRanges_ex1 and the GRangesList derived from the ASCAT data containing 3 samples i.e. sample_list_GRanges_ex2. We can now use this GRangesList as the input to RaggedExperiment.

Using GRangesList where ASCAT data only has 1 sample:

ragexp_2 <- RaggedExperiment(
    sample_list_GRanges_ex1,
    colData = colDat_1
)

ragexp_2

Using GRangesList where ASCAT data only has multiple samples:

colDat_3 = DataFrame(id = 1:3)

ragexp_3 <- RaggedExperiment(
    sample_list_GRanges_ex2,
    colData = colDat_3
)

ragexp_3

We can also use the GRangesList produced using the GRangesList function:

ragexp_4  <- RaggedExperiment(
    sample_list_GRanges_ex3,
    colData = colDat_1
)

ragexp_4

Downstream Analysis

Now that we have the ASCAT data converted to RaggedExperiment objects we can use the *Assay functions that are described in the RaggedExperiment vignette. These functions provide several different functions for representing ranged data in a rectangular matrix. They make it easy to find genomic segments shared/not shared between each sample considered and provide the corresponding allele-specific copy number calls for each sample across each segment.

Session Information

sessionInfo()

Bioconductor/RaggedExperiment documentation built on Oct. 31, 2024, 7:02 a.m.