fgga: Factor Graph Cross-Ontology Annotation model

In fspetale/fgga: Hierarchical ensemble method based on factor graph

Factor Graph Cross-Ontology Annotation model

Description

A hierarchical graph-based machine learning model for the consistent GO, PO, ZFA, HPO annotation of protein coding genes.

Usage

fgga(graphOnto, tableOntoTerms, dxCharacterized, dxTestCharacterized,
    kFold, kernelSVM, tmax, epsilon)

Arguments

graphOnto	A graphNEL graph with ‘m’ Ontology node labels.
tableOntoTerms	A binary matrix with ‘n’ proteins (rows) by ‘m’ Ontology node labels (columns).
dxCharacterized	A data frame with ‘n’ proteins (rows) by ‘f’ features (columns).
dxTestCharacterized	A data frame with ‘k’ proteins (rows) by ‘f’ features (columns).
kFold	An integer for the number of folds.
kernelSVM	The kernel used to calculate the variance (default: radial).
tmax	An integer indicating the maximum number of iterations (default: 200).
epsilon	A real value less than 1 that represents the convergence criteria (default: 0.001).

Details

The FGGA model is built in two main steps. In the first step, a core Factor Graph (FG) modeling hidden Ontology-term predictions and relationships is created. In the second step, the FG is enriched with nodes modeling observable Ontology-term predictions issued by binary SVM classifiers. In addition, probabilistic constraints modeling learning gaps between hidden and observable Ontology-term predictions are introduced. These gaps are assumed to be independent among Ontology-terms, locally additive with respect to observed predictions, and zero-mean Gaussian. FGGA predictions are issued by the native iterative message passing algorithm of factor graphs.

Value

A named matrix with ‘k’ protein coding genes (rows) by ‘m’ cross-Ontology node labels (columns) where each element indicates a probabilistic prediction value.

Author(s)

Flavio E. Spetale and Elizabeth Tapia <[e-mail]>

References

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

Spetale Flavio E., Arce D., Krsticevic F., Bulacio P. and Tapia E. “Consistent prediction of GO protein localization”. Scientific Report 7787(8), 2018

See Also

fgga2bipartite, sumProduct, svmOnto

Examples

data(CfData)

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

dxCfTestCharacterized <- CfData[["dxCf"]][CfData[["indexGO"]]$indexTest[1:2], ]

myTableGO <- CfData[["tableCfGO"]][
                    CfData[["indexGO"]]$indexTrain[1:300], ]

dataTrain <- CfData[["dxCf"]][
                    CfData[["indexGO"]]$indexTrain[1:300], ]

fggaResults <- fgga(graphOnto = mygraphGO,
                tableOntoTerms = myTableGO, dxCharacterized = dataTrain,
                dxTestCharacterized = dxCfTestCharacterized, kFold = 2,
                tmax = 50, epsilon = 0.05)

fspetale/fgga documentation built on Jan. 29, 2024, 6:53 p.m.