geva.quantiles: GEVA Quantiles Detection

In sbcblab/geva: Gene Expression Variation Analysis (GEVA)

Description

Calculates the quantiles of a SVTable.

Usage

geva.quantiles(
  sv,
  quantile.method = options.quantiles,
  initial.thresholds = c(S = NA_real_, V = NA_real_),
  nq.s = 3L,
  nq.v = 2L,
  comb.score.fn = prod,
  ...
)

options.quantiles
# c("range.slice", "proportional", "density", "k.max.sd", 
#   "custom")

Arguments

sv	a SVTable object (usually GEVASummary)
quantile.method	character, method to detect the initial quantile thresholds. Ignored if initial.thresholds is specified with no NA elements
initial.thresholds	named numeric vector with the threshold that delimits the initial quantile
nq.s	integer, number of quantiles in S-axis (experimental, see ‘Note’)
nq.v	integer, number of quantiles in V-axis (experimental, see ‘Note’)
comb.score.fn	function applied to merge S and V score columns into a single column. The function must require only one argument of numeric vector type and return a single numeric value. Examples include prod or mean
...	additional arguments include: qslice : numeric (0 to 1), the axis fraction used by "range.slice" and "density" methods (see 'Details'). Default is 0.25 k : integer, neighbor points used by "density" and "k.max.sd" methods (see 'Details'). Default is 16 verbose : logical, whether to print the current progress. Default is TRUE

Details

The quantile.method defines how the initial quantile (usually the one at the bottom center) is calculated. Each method has a specific way to estimate the first spatial delimiter, as described below:

"range.slice" (default)

Separation is set at the nearest point to a fraction of the spatial range. This fraction can be specified by the qslice optional argument (numeric, default is 0.25, or 25%);

"density"

Separation is set at the point with the most proportional density by k neighbor points to its current spatial fraction. This method uses the optional arguments qslice (numeric, default is 0.25, or 25%) for the desired spatial fraction, and k (numeric, default is 16) for the number of neighbor points;

"k.max.sd"

Separation is set at the point with the greatest standard deviation of distance to its k neighbor points. The number of neighbor points can be specified by the k optional argument (numeric, default is 16);

"proportional"

Separation is set at the exact axis division so that all quantiles have the size;

"custom"

Uses the values specified in the initial.thresholds argument.

A custom initial separation point can be specified in the initial.thresholds as a numeric vector of two elements, where the first element refers to S axis and the second, to V axis. If one of the elements is NA, the initial quantile is calculated for that axis only. If both values are not NA, the quantile separation method is ignored and automatically set to "custom".

The nq.s and nq.v arguments determine the number of quantiles for the S and V axes, respectively. These parameters can be used to increase the number of possible partitions in the SV space, but their applicability is currently being tested (see ‘Note’).

The comb.score.fn is a function applied to the partial scores for each SV point to combine them into a single value. The result value is defined as the "quantile score" for a SV point. The function is applied iteratively to two-element numeric vectors.

Note

Customizing the number of quantiles by nq.s and nq.v is a experimental feature and the remaining analysis steps are mostly based on the default parameters for these arguments. Tests are being conducted to determine this feature's applicability for the next releases.

See Also

geva.cluster

Other geva.cluster: geva.cluster(), geva.dcluster(), geva.hcluster()

Examples

## Quantile detection from a randomly generated input 

# Preparing the data
ginput <- geva.ideal.example()      # Generates a random input example
gsummary <- geva.summarize(ginput)  # Summarizes with the default parameters

# Default usage
gquants <- geva.quantiles(gsummary) # Detects the quantiles
plot(gquants)                       # Plots the quantiles

# Custom initial delimiters
gquants <- geva.quantiles(gsummary,
                          initial.thresholds = c(S=1.00, V=0.5))
plot(gquants)                       # Plots the quantiles

# Quantile detection using densities
gquants <- geva.quantiles(gsummary, quantile.method = 'density')
plot(gquants)                       # Plots the quantiles

sbcblab/geva documentation built on March 15, 2021, 10:08 p.m.