gRxCluster-object: gRxCluster object

Description Details Author(s)

Description

Overview of the result of gRxCluster(...)

Details

The object returned is a GRanges object.

If the object is x, seqnames(x) and ranges(x) slots demarcate the clusters discovered. There will be one element for each cluster (aka ‘clump’) discovered.

Using the default argument pruneFun=prune.loglik or pruneFun=noprune, mcols(x) will have these columns:

value1 and value2

are the counts of the two classes of insertion sites for the clusters of object x

clump.id

numbers each cluster.

If the user supplies a custom pruneFun, it should return a GRanges with those columns and one element for each unique clump.id. The column target.min has the smallest nominal False Discoveries Expected for each cluster and is added to (or replaces) the mcols(x) produced by the argument supplied as pruneFun.

metadata(x) will include these components:

criticalValues

A list object such as supplied by critVal.target whose elements each give the cutpoints to be used for a window with k sites. attributes(metadata(object)$criticalValues[[i]]) will contain elements

fdr

with dimension c(k+1,4) of target false discovery expectations and and the one-sided p-values

target

the target for false discovery which sometimes is specified a priori and sometimes results from calculation

n

an upper bound on the number of windows to screen, if this number is needed.

In some cases, an attribute is attached to metadata(object)$criticalValues, see critVal.power for an example.

kvals

the number of sites, k, to include in a window

perm_cluster_best

a list whose canonical element is a vector of values like x$target.min obtained from a permutation of the class indicators

summary_matrix

a matrix giving the start, end, depth, and counts in each class for every cluster and depth in sequential order

call

the call invoking gRxCluster

which may include some arguments added by default.

Author(s)

Charles Berry ccberry@ucsd.edu


geneRxCluster documentation built on Nov. 8, 2020, 4:56 p.m.