clusterCTSS: Clustering CTSSs into tag clusters (TCs)

Description

Clusters individual CAGE transcription start sites (CTSSs) along the genome into tag clusters using specified "ab initio" method, or assigns them to predefined genomic regions.

Usage

clusterCTSS(object, threshold = 1, nrPassThreshold = 1, thresholdIsTpm = TRUE, method = "distclu", maxDist = 20,  removeSingletons = FALSE, keepSingletonsAbove = Inf,  minStability = 1, maxLength = 500,  reduceToNonoverlapping = TRUE, customClusters = NULL,  useMulticore = FALSE, nrCores = NULL)

Arguments

object

A CAGEset object

threshold, nrPassThreshold

Only CTSSs with signal >= threshold in >= nrPassThreshold experiments will be used for clustering and will contribute towards total signal of the cluster.

thresholdIsTpm

Logical, is threshold raw tag count value (FALSE) or normalized signal (TRUE)

method

Method to be used for clustering. Can be one of the "distclu", "paraclu" or "custom". See Details.

maxDist

Maximal distance between two neighbouring CTSSs for them to be part of the same cluster. Used only when method = "distclu", otherwise ignored.

removeSingletons

Logical, should tag clusters containing only one CTSS be removed. Ignored when method = "custom".

keepSingletonsAbove

Controls which singleton tag clusters will be removed. When removeSingletons = TRUE, only singletons with signal < keepSingletonsAbove will be removed. Useful to prevent removing highly supported singleton tag clusters. Default value Inf results in removing all singleton TCs when removeSingletons = TRUE. Ignored when removeSingletons = FALSE or method = "custom".

minStability

Minimal stability of the cluster, where stability is defined as ratio between maximal and minimal density value for which this cluster is maximal scoring. For definition of stability refer to Frith et al., Genome Research, 2007. Clusters with stability < minStability will be discarded. Used only when method = "paraclu", otherwise ignored.

maxLength

Maximal length of cluster in base-pairs. Clusters with length > maxLength will be discarded. Ignored when method = "custom".

reduceToNonoverlapping

Logical, should smaller clusters contained within bigger cluster be removed to make a final set of tag clusters non-overlapping. Used only when method = "paraclu". See Details.

customClusters

Genomic coordinates of predefined regions to be used to segment the CTSSs. It has to be a data.frame with following columns: chr (chromosome name), start (0-based start coordinate), end (end coordinate), strand (either "+", or "-"). Used only when method = "custom".

useMulticore

Logical, should multicore be used. useMulticore = TRUE is supported only on Unix-like platforms.

nrCores

Number of cores to use when useMulticore = TRUE. Default value NULL uses all detected cores.

Value

The slots clusteringMethod, filteredCTSSidx and tagClusters of the provided CAGEset object will be occupied by the information on method used for clustering, CTSSs included in the clusters and list of tag clusters per CAGE experiment, respectively. To retrieve tag clusters for individual CAGE dataset use tagClusters function.

Details

Two "ab initio" methods for clustering TSSs along the genome are supported: "distclu" and "paraclu". "distclu" is an implementation of simple distance-based clustering of data attached to sequences, where two neighbouring TSSs are joined together if they are closer than some specified distance. "paraclu" is an implementation of Paraclu algorithm for parametric clustering of data attached to sequences developed by M. Frith (Frith et al., Genome Research, 2007, http://www.cbrc.jp/paraclu/). Since Paraclu finds clusters within clusters (unlike distclu), additional parameters (removeSingletons, keepSingletonsAbove, minStability, maxLength and reduceToNonoverlapping) can be specified to simplify the output by discarding too small (singletons) or too big clusters, and to reduce the clusters to a final set of non-overlapping clusters. Clustering is done for every CAGE dataset within CAGEset object separatelly, resulting in a different set of tag clusters for every CAGE dataset. TCs from different datasets can further be aggregated into a single referent set of consensus clusters by calling aggregateTagClusters function.

References

Frith et al. (2007) A code for transcription initiation in mammalian genomes, Genome Research 18(1):1-12, (http://www.cbrc.jp/paraclu/).

Examples

Run this code

load(system.file("data", "exampleCAGEset.RData", package="CAGEr"))

clusterCTSS(object = exampleCAGEset, threshold = 50, thresholdIsTpm = TRUE,
nrPassThreshold = 1, method = "distclu", maxDist = 20, 
removeSingletons = TRUE, keepSingletonsAbove = 100)

Run the code above in your browser using DataLab