varianceGO {fgga}R Documentation

The variance of the gaussian learning noise at individual GO-terms

Description

varianceGO estimates the variance of gaussian distributions modeling the additive learning noise that corrupts ideal GO-term predictions.

Usage

varianceGO(tableGOs, dxCharacterized, kFold, graphGO, rootNode,
            kernelSVM = "radial")

Arguments

tableGOs

A binary matrix with ‘n’ protein coding genes (rows) by ‘m’ GO node labels (columns).

dxCharacterized

A data frame with ‘n’ protein coding genes (rows) by ‘f’ features (columns).

kFold

An integer for the number of folds.

graphGO

A graphNEL graph with ‘m’ GO node labels.

rootNode

A character indicating the root of the graph.

kernelSVM

The kernel used to calculate the variance (default: radial).

Details

Under the assumption of symmetrical (Gaussian) conditional probability distributions for observable variable node predictions y_i over a hidden variable node annotations x_i, variances η_i can be estimated using a validation dataset of positively annotated samples.

Value

A vector named with the variance of each GO node.

Author(s)

Flavio E. Spetale and Javier Murillo <spetale@cifasis-conicet.gov.ar>

References

Spetale FE, Tapia E, Krsticevic F, Roda F, Bulacio P (2016). A Factor Graph Approach to Automated GO Annotation. PLOS ONE 11(1): e0146986

Examples

data(CfData)

mygraphGO <- as(CfData[["graphCfGO"]], "graphNEL")

rootGO <- leaves(mygraphGO, "in")

mygraphGO <- subGraph(c("GO:0140110", "GO:0098772", "GO:0003674"), mygraphGO)

myTableGO <- CfData[["tableCfGO"]][
                    CfData[["indexGO"]]$indexTrain,
                    c("GO:0140110", "GO:0098772", "GO:0003674")]

varianceGOs <- varianceGO(tableGOs = myTableGO,
                            dxCharacterized = CfData[["dxCf"]],
                            kFold = 2, graphGO = mygraphGO,
                            rootNode = rootGO, kernelSVM = "radial")

[Package fgga version 1.0.0 Index]