--- title: 'Overview of katdetectr' date: "`r BiocStyle::doc_date()`" author: "Daan Hazelaar and Job van Riet" vignette: > %\VignetteIndexEntry{Overview_katdetectr} %\VignetteEngine{knitr::rmarkdown} %\VignetteEncoding{UTF-8} package: 'katdetectr' output: BiocStyle::html_document: toc: TRUE --- # Installation To install this package, start R (version "4.2") and enter: ```{r eval=FALSE} # Install via BioConductor if (!require("BiocManager", quietly = TRUE)) install.packages("BiocManager") BiocManager::install("katdetectr") # or the latest version BiocManager::install("katdetectr", version = "devel") # or from github devtools::install_git("https://github.com/ErasmusMC-CCBC/katdetectr") ``` ```{r} library(katdetectr) ``` # Introduction `katdetectr` is an *R* package for the detection, characterization and visualization of localized hypermutated regions, often referred to as *kataegis*. The general workflow of `katdetectr` can be summarized as follows: 1. Import of genomic variants; VCF, MAF or VRanges objects. 2. Detection of kataegis foci. 3. Visualization of segmentation and kataegis foci. Please see the [Application Note](https://www.biorxiv.org/content/10.1101/2022.07.11.499364v1) (under submission) for additional background and details of `katdetectr`. The application note also contains a section regarding the performance of `katdetectr` and other kataegis detection packages: [maftools](https://bioconductor.org/packages/release/bioc/html/maftools.html), [ClusteredMutations](https://cran.r-project.org/web/packages/ClusteredMutations/index.html), [SeqKat](https://cran.r-project.org/web/packages/SeqKat/index.html), [kataegis](https://github.com/flosalbizziae/kataegis), and [SigProfilerClusters](https://github.com/AlexandrovLab/SigProfilerClusters). We have made `katdetectr` available on *BioConductor* as this insures reliability, and operability on common operating systems (Linux, Mac, and Windows). We designed `katdetectr` such that it fits well in the *BioConductor* ecosystem which allows `katdetectr` to be used easily in combination with other *BioConductor* packages and analysis pipelines. Below, the `katdetectr` workflow is performed in a step-by-step manner on publicly-available datasets that are included within this package. # Importing genomic variants Genomic variants from multiple common data-formats (VCF/MAF and VRanges objects) can be imported into `katdetectr`. ```{r, label = 'Importing genomic variants', eval = TRUE} # Genomic variants stored within the VCF format. pathToVCF <- system.file(package = "katdetectr", "extdata/CPTAC_Breast.vcf") # Genomic variants stored within the MAF format. pathToMAF <- system.file(package = "katdetectr", "extdata/APL_primary.maf") # In addition, we can generate synthetic genomic variants including kataegis foci using generateSyntheticData(). # This functions returns a VRanges object. syntheticData <- generateSyntheticData(nBackgroundVariants = 2500, nKataegisFoci = 1) ``` # Detection of kataegis foci Using `detectKataegis()`, we can employ changepoint detection to detect distinct clusters of varying intermutation distance (IMD), mutation rate and size. Imported genomic variant data can contain either single or multiple samples, in the latter case records can be aggregated by setting `aggregateRecords = TRUE`. Overlapping genomic variants (e.g., an InDel and SNV) are reduced into a single record. From the genomic variants data the IMD is calculated. Following, changepoint analysis is performed on the IMD of the genomic variants which results in segments. Lastly, a segment is labelled as *kataegis foci* if the segment fits the following parameters: ` minSizeKataegis = 6` and `IMDcutoff = 1000`. ```{r, label = 'Detection of kataegis foci', eval = TRUE} # Detect kataegis foci within the given VCF file. kdVCF <- detectKataegis(genomicVariants = pathToVCF) # # Detect kataegis foci within the given MAF file. # As this file contains multiple samples, we set aggregateRecords = TRUE. kdMAF <- detectKataegis(genomicVariants = pathToMAF, aggregateRecords = TRUE) # Detect kataegis foci within the synthetic data. kdSynthetic <- detectKataegis(genomicVariants = syntheticData) ``` All relevant input and subsequent results are stored within `KatDetect` objects. Using `summary()`, `show()` and/or `print()`, we can generate overviews of these `KatDetect` object(s). ```{r, label = 'Overview of KatDetect objects', eval = TRUE} summary(kdVCF) print(kdVCF) show(kdVCF) # Or simply: kdVCF ``` Underlying data can be retrieved from a `KatDetect` objects using the following getter functions: 1. `getGenomicVariants()` returns: VRanges object. Processed genomic variants used as input for changepoint detection. This VRanges contains the genomic location, IMD, and kataegis status of each genomic variant 2. `getSegments()` returns: GRanges object. Contains the segments as derived from changepoint detection. This Granges contains the genomic location, total number of variants, mean IMD and, mutation rate of each segment. 3. `getKataegisFoci()` returns: GRanges object. Contains all segments designated as putative kataegis foci according the the specified parameters (minSizeKataegis and IMDcutoff). This Granges contains the genomic location, total number of variants and mean IMD of each putative kataegis foci 4. `getInfo()` returns: List object. Contains supplementary information including used parameter settings. ```{r, label = 'Retrieving information from KatDetect objects', eval = TRUE} getGenomicVariants(kdVCF) getSegments(kdVCF) getKataegisFoci(kdVCF) getInfo(kdVCF) ``` # Visualization of segmentation and kataegis foci Per sample, we can visualize the IMD, detected segments and putative kataegis foci as a rainfall plot. In addition, this allows for a per-chromosome approach which can highlight the putative kataegis foci. ```{r, label = 'Visualisation of KatDetect objects', eval = TRUE} rainfallPlot(kdVCF) # With showSegmentation, the detected segments (changepoints) as visualized with their mean IMD. rainfallPlot(kdMAF, showSegmentation = TRUE) # With showSequence, we can display specific chromosomes or all chromosomes in which a putative kataegis foci has been detected. rainfallPlot(kdSynthetic, showKataegis = TRUE, showSegmentation = TRUE, showSequence = "Kataegis") ``` # More parameter settings `katdetectr` has been implemented flexibly which allows its users to detect clustered mutations of different classes. The historical definition of kataegis foci, [postulated by Alexandrov et al.*](https://www.nature.com/articles/nature12477) is a segment harboring ≥6 variants and has a mean IMD ≤1000bp. However, these parameters can be set differently in `detectKataegis()`. For example, another type of clustered mutations are called Omikli: a segment harboring 2 or 3 variants with mean IMD = m Note that we did not evaluate the performance of `katdetectr` in regards to detecting other types of clusters. The following is just to show you how to change the parameters of `detectKataegis()` if you want to use `katdetectr` for detecting these types of clusters. ```{r} # detect putative Omikli, size 3 and mean IMD = 500 kdSyntheticOmikli500 <- detectKataegis(genomicVariants = syntheticData, minSizeKataegis = 3, IMDcutoff = 500) # detect putative Omikli, size 4 and mean IMD = 1000 kdSyntheticOmikli500 <- detectKataegis(genomicVariants = syntheticData, minSizeKataegis = 4, IMDcutoff = 1000) ``` We tested katdetectr with multiple parameter settings (test.stat, penalty, pen.value, minseglen) in order to obtain the highest performance in regards to kataegis classification. The best combination of parameters have been set as the default values. We recommend using these parameter settings! For more information regarding the these parameters for changepoint analysis see [Killick2014](https://www.jstatsoft.org/article/view/v058i03) # Custom function for IMD cutoff value. As show previously, the IMD cutoff value can be set by the user in order to identify different types of mutation clusters. However, it is also possible to define a custom function that determines a IMD cutoff value for each detected segment. This flexibility allow you to define your own kataegis definition while still using the `katdetectr` framework. Below we show how to implement a custom function for the IMD cutoff value. The function below comes from the work of [Pan-Cancer Analysis of Whole Genomes Consortium](https://www.nature.com/articles/s41586-020-1969-6#citeas) ```{r} # function for modeling sample mutation rate modelSampleRate <- function(IMDs){ lambda <- log(2) / median(IMDs) return(lambda) } # function for calculating the nth root of x nthroot = function(x, n){ y <- x^(1 / n) return(y) } # Function that defines the IMD cutoff specific for each segment # Within this function you can use all variables available in the slots: genomicVariants and segments IMDcutoffFun <- function(genomicVariants, segments){ IMDs <- genomicVariants$IMD totalVariants <- segments$totalVariants width <- segments |> dplyr::as_tibble() |> dplyr::pull(width) sampleRate <- modelSampleRate(IMDs) IMDcutoff <- -log(1 - nthroot(0.01 / width, ifelse(totalVariants != 0, totalVariants - 1, 1))) / sampleRate IMDcutoff <- replace(IMDcutoff, IMDcutoff > 1000, 1000) return(IMDcutoff) } kdCustom <- detectKataegis(syntheticData, IMDcutoff = IMDcutoffFun) ``` # Analyzing non standard sequences The human autosomes and sex chromosomes from the reference genome hg19 and hg38 come implemented in `katdetectr`. However, other (non standard) sequences can be analyzed if the correct arguments are provided. You have to provide a dataframe that contains the length of all the sequences you want to analyze. ```{r} # generate data that contains non standard sequences syndata1 <- generateSyntheticData(seqnames = c("chr1_gl000191_random", "chr4_ctg9_hap1")) syndata2 <- generateSyntheticData(seqnames = "chr1") syndata <- suppressWarnings(c(syndata1, syndata2)) # construct a dataframe that contains the length of the sequences # each column name (name of the sequence) sequenceLength = data.frame( chr1 = 249250621, chr1_gl000191_random = 106433, chr4_ctg9_hap1 = 590426 ) # provide the dataframe with the sequence lengths using the refSeq argument kdNonStandard <- detectKataegis(genomicVariants = syndata, refSeq = sequenceLength) ``` # Session Information ```{r, label = 'Session Information', eval = TRUE} utils::sessionInfo() ```