| Title: | Inferring functionally related proteins using protein interaction networks |
|---|---|
| Description: | Interactions between proteins occur in many, if not most, biological processes. Most proteins perform their functions in networks associated with other proteins and other biomolecules. This fact has motivated the development of a variety of experimental methods for the identification of protein interactions. This variety has in turn ushered in the development of numerous different computational approaches for modeling and predicting protein interactions. Sometimes an experiment is aimed at identifying proteins closely related to some interesting proteins. A network based statistical learning method is used to infer the putative functions of proteins from the known functions of its neighboring proteins on a PPI network. This package identifies such proteins often involved in the same or similar biological functions. |
| Authors: | Dongmin Jung, Xijin Ge |
| Maintainer: | Dongmin Jung <[email protected]> |
| License: | Artistic-2.0 |
| Version: | 1.31.0 |
| Built: | 2024-07-05 05:18:42 UTC |
| Source: | https://github.com/bioc/PPInfer |
Interactions between proteins occur in many, if not most, biological processes. Most proteins perform their functions in networks associated with other proteins and other biomolecules. This fact has motivated the development of a variety of experimental methods for the identification of protein interactions. This variety has in turn ushered in the development of numerous different computational approaches for modeling and predicting protein interactions. Sometimes an experiment is aimed at identifying proteins closely related to some interesting proteins. A network based statistical learning method is used to infer the putative functions of proteins from the known functions of its neighboring proteins on a PPI network. This package identifies such proteins often involved in the same or similar biological functions.
The DESCRIPTION file:
| Package: | PPInfer |
| Type: | Package |
| Title: | Inferring functionally related proteins using protein interaction networks |
| Description: | Interactions between proteins occur in many, if not most, biological processes. Most proteins perform their functions in networks associated with other proteins and other biomolecules. This fact has motivated the development of a variety of experimental methods for the identification of protein interactions. This variety has in turn ushered in the development of numerous different computational approaches for modeling and predicting protein interactions. Sometimes an experiment is aimed at identifying proteins closely related to some interesting proteins. A network based statistical learning method is used to infer the putative functions of proteins from the known functions of its neighboring proteins on a PPI network. This package identifies such proteins often involved in the same or similar biological functions. |
| Version: | 1.31.0 |
| Date: | 2022-11-17 |
| Author: | Dongmin Jung, Xijin Ge |
| Maintainer: | Dongmin Jung <[email protected]> |
| Depends: | biomaRt, fgsea, kernlab, ggplot2, igraph, STRINGdb, yeastExpData |
| Imports: | httr, grDevices, graphics, stats, utils |
| Suggests: | |
| License: | Artistic-2.0 |
| biocViews: | Software, StatisticalMethod, Network, GraphAndNetwork, GeneSetEnrichment, NetworkEnrichment, Pathways |
| NeedsCompilation: | no |
| Repository: | https://bioc.r-universe.dev |
| RemoteUrl: | https://github.com/bioc/PPInfer |
| RemoteRef: | HEAD |
| RemoteSha: | 789e93006aa2cbbf5a5667fd1540048344ce7e2c |
Index of help topics:
GSEA.barplot Visualize the gene set enrichment analysis
ORA Over-representation Analysis
ORA.barplot Visualize the over-representation analysis
PPInfer-package Inferring functionally related proteins using
protein interaction networks
enrich.net Visualize network for the functional enrichment
analysis
net.infer Inferring functionally related proteins using
networks
net.infer.ST Inferring functionally related proteins with
self training
net.kernel Kernel matrix for a graph
ppi.infer.human Inferring functionally related proteins using
protein networks for human
ppi.infer.mouse Inferring functionally related proteins using
protein networks for mouse
self.train.kernel Self training for a kernel matrix
Further information is available in the following vignettes:
PPInfer |
User manual (source, pdf) |
Dongmin Jung, Xijin Ge
Maintainer: Dongmin Jung <[email protected]>
The connection between nodes depends on the proportion of overlapping genes between two categories.
enrich.net(x, gene.set, node.id, node.name = node.id, pvalue, n = 50, numChar = NULL, pvalue.cutoff = 0.05, edge.cutoff = 0.05, degree.cutoff = 0, edge.width = function(x) {10*x^2}, node.size = function(x) {2.5*log10(x)}, group = FALSE, group.color = c('red', 'green'), group.shape = c('circle', 'square'), legend.parameter = list('topright'), show.legend = TRUE, ...)enrich.net(x, gene.set, node.id, node.name = node.id, pvalue, n = 50, numChar = NULL, pvalue.cutoff = 0.05, edge.cutoff = 0.05, degree.cutoff = 0, edge.width = function(x) {10*x^2}, node.size = function(x) {2.5*log10(x)}, group = FALSE, group.color = c('red', 'green'), group.shape = c('circle', 'square'), legend.parameter = list('topright'), show.legend = TRUE, ...)
x |
a result with category and p-value of gene sets |
gene.set |
gene sets which is already used for functional enrichment |
node.id |
name of gene sets |
node.name |
label of nodes in the network (default: node.id) |
pvalue |
pvalues for categories |
n |
number of top categories (default: 50) |
numChar |
the maximal number of characters of the label of gene sets |
pvalue.cutoff |
nodes with p-values which are greater than pvalue.cutoff are removed (default: 0.05) |
edge.cutoff |
edges with the proportion which is less than edge.cutoff are removed (default: 0.05) |
degree.cutoff |
nodes with the degrees which are less than degree.cutoff are removed (default: 0) |
edge.width |
width of edges |
node.size |
size of nodes |
group |
variable for group |
group.color |
color for group (default: red and green for 2 groups) |
group.shape |
shape for group (default: circle and square for 2 groups) |
legend.parameter |
list of parametres for the legend |
show.legend |
show the legend (default: TRUE) |
... |
additional parameters for the igraph |
plot for the network. The size of nodes is proportional to the size of gene sets. The more significant categories are, the less transparent their nodes are.
Dongmin Jung, Xijin Ge
Yu G, Wang L, Yan G and He Q (2015). "DOSE: an R/Bioconductor package for Disease Ontology Semantic and Enrichment analysis." Bioinformatics, 31(4), pp. 608-609.
igraph
data(examplePathways) data(exampleRanks) set.seed(1) result.GSEA <- fgsea(examplePathways, exampleRanks, nperm = 1000) enrich.net(result.GSEA, examplePathways, node.id = 'pathway', pvalue = 'pval', edge.cutoff = 0.6, degree.cutoff = 1, n = 50, vertex.label.cex = 0.75, show.legend = FALSE, edge.width = function(x) {5*sqrt(x)}, layout = igraph::layout.kamada.kawai)data(examplePathways) data(exampleRanks) set.seed(1) result.GSEA <- fgsea(examplePathways, exampleRanks, nperm = 1000) enrich.net(result.GSEA, examplePathways, node.id = 'pathway', pvalue = 'pval', edge.cutoff = 0.6, degree.cutoff = 1, n = 50, vertex.label.cex = 0.75, show.legend = FALSE, edge.width = function(x) {5*sqrt(x)}, layout = igraph::layout.kamada.kawai)
For the functional enrichment analysis, we can visualize the result from the gene set enrichment analysis.
GSEA.barplot(object, category, score, pvalue, top = 10, sort = NULL, decreasing = FALSE, numChar = NULL, title = NULL, transparency = 0.5, plot = TRUE)GSEA.barplot(object, category, score, pvalue, top = 10, sort = NULL, decreasing = FALSE, numChar = NULL, title = NULL, transparency = 0.5, plot = TRUE)
object |
a table with category, enrichment score and p-value of gene sets |
category |
name of gene sets |
score |
enrichment score |
pvalue |
p-value of gene sets |
top |
the number of top categories (default: 10) |
sort |
a variable used for sorting data |
decreasing |
logical indicating whether ascending or descending order (default: FALSE) |
numChar |
the maximal number of characters of the name of gene sets |
title |
title for the plot |
transparency |
transparency (default: 0.5) |
plot |
return plot when plot is true, otherwise return table (default: TRUE) |
GSEA barplot
Dongmin Jung, Xijin Ge
Yu G, Wang L, Yan G and He Q (2015). "DOSE: an R/Bioconductor package for Disease Ontology Semantic and Enrichment analysis." Bioinformatics, 31(4), pp. 608-609.
ggplot2
data(examplePathways) data(exampleRanks) set.seed(1) result.GSEA <- fgsea(examplePathways, exampleRanks, nperm = 1000) GSEA.barplot(result.GSEA, category = 'pathway', score = 'NES', pvalue = 'pval', sort = 'NES', decreasing = TRUE)data(examplePathways) data(exampleRanks) set.seed(1) result.GSEA <- fgsea(examplePathways, exampleRanks, nperm = 1000) GSEA.barplot(result.GSEA, category = 'pathway', score = 'NES', pvalue = 'pval', sort = 'NES', decreasing = TRUE)
Proteins can be classified by using networks to identify functionally closely related proteins.
net.infer(target, kernel, top = NULL, cross = 0, C = 1, nu = 0.2, epsilon = 0.1, cache1 = 40, tol1 = 0.001, shrinking1 = TRUE, cache2 = 40, tol2 = 0.001, shrinking2 = TRUE)net.infer(target, kernel, top = NULL, cross = 0, C = 1, nu = 0.2, epsilon = 0.1, cache1 = 40, tol1 = 0.001, shrinking1 = TRUE, cache2 = 40, tol2 = 0.001, shrinking2 = TRUE)
target |
set of interesting proteins or target class |
kernel |
the regularized Laplacian matrix for a graph |
top |
number of top proteins most closely related to target class (default: all proteins except for target and pseudo-absence class) |
cross |
if a integer value k>0 is specified, a k-fold cross validation on the training data is performed to assess the quality of the model |
C |
cost of constraints violation for SVM (default: 1) |
nu |
The nu parameter for OCSVM (default: 0.2) |
epsilon |
epsilon in the insensitive-loss function for OCSVM (default: 0.1) |
cache1 |
cache memory in MB for OCSVM (default: 40) |
tol1 |
tolerance of termination criterion for OCSVM (default: 0.001) |
shrinking1 |
option whether to use the shrinking-heuristics for OCSVM (default: TRUE) |
cache2 |
cache memory in MB for SVM (default: 40) |
tol2 |
tolerance of termination criterion for SVM (default: 0.001) |
shrinking2 |
option whether to use the shrinking-heuristics for SVM (default: TRUE) |
list |
list of a target class used in the model |
error |
training error |
CVerror |
cross validation error, (when cross > 0) |
top |
top proteins |
score |
decision values for top proteins |
Dongmin Jung, Xijin Ge
Senay, S. D. et al. (2013). Novel three-step pseudo-absence selection technique for improved species distribution modelling. PLOS ONE. 8(8), e71218.
ksvm
# example 1 ## Not run: string.db.9606 <- STRINGdb$new(version = '11', species = 9606, score_threshold = 999) string.db.9606.graph <- string.db.9606$get_graph() K.9606 <- net.kernel(string.db.9606.graph) rownames(K.9606) <- substring(rownames(K.9606), 6) colnames(K.9606) <- substring(colnames(K.9606), 6) target <- colnames(K.9606)[1:100] infer <- net.infer(target, K.9606, 10) ## End(Not run) # example 2 data(litG) litG <- igraph.from.graphNEL(litG) sg <- decompose(litG, min.vertices = 50) sg <- sg[[1]] K <- net.kernel(sg) litG.infer <- net.infer(names(V(sg))[1:10], K, top=20)# example 1 ## Not run: string.db.9606 <- STRINGdb$new(version = '11', species = 9606, score_threshold = 999) string.db.9606.graph <- string.db.9606$get_graph() K.9606 <- net.kernel(string.db.9606.graph) rownames(K.9606) <- substring(rownames(K.9606), 6) colnames(K.9606) <- substring(colnames(K.9606), 6) target <- colnames(K.9606)[1:100] infer <- net.infer(target, K.9606, 10) ## End(Not run) # example 2 data(litG) litG <- igraph.from.graphNEL(litG) sg <- decompose(litG, min.vertices = 50) sg <- sg[[1]] K <- net.kernel(sg) litG.infer <- net.infer(names(V(sg))[1:10], K, top=20)
This function is the self-training version of net.infer. The function net.infer is the special case of net.infer.ST where a single iteration is conducted.
net.infer.ST(target, kernel, top = NULL, C = 1, nu = 0.2, epsilon = 0.1, cache1 = 40, tol1 = 0.001, shrinking1 = TRUE, cache2 = 40, tol2 = 0.001, shrinking2 = TRUE, thrConf = 0.9, maxIts = 10, percFull = 1, verbose = FALSE)net.infer.ST(target, kernel, top = NULL, C = 1, nu = 0.2, epsilon = 0.1, cache1 = 40, tol1 = 0.001, shrinking1 = TRUE, cache2 = 40, tol2 = 0.001, shrinking2 = TRUE, thrConf = 0.9, maxIts = 10, percFull = 1, verbose = FALSE)
target |
set of interesting proteins or target class |
kernel |
the regularized Laplacian matrix for a graph |
top |
number of top proteins most closely related to target class (default: all proteins except for target and pseudo-absence class) |
C |
cost of constraints violation for SVM (default: 1) |
nu |
The nu parameter for OCSVM (default: 0.2) |
epsilon |
epsilon in the insensitive-loss function for OCSVM (default: 0.1) |
cache1 |
cache memory in MB for OCSVM (default: 40) |
tol1 |
tolerance of termination criterion for OCSVM (default: 0.001) |
shrinking1 |
option whether to use the shrinking-heuristics for OCSVM (default: TRUE) |
cache2 |
cache memory in MB for SVM (default: 40) |
tol2 |
tolerance of termination criterion for SVM (default: 0.001) |
shrinking2 |
option whether to use the shrinking-heuristics for SVM (default: TRUE) |
thrConf |
A number between 0 and 1, indicating the required classification confidence for an unlabelled case to be added to the labelled data set with the label predicted predicted by the classification algorithm (default: 0.9) |
maxIts |
The maximum number of iterations of the self-training process (default: 10) |
percFull |
A number between 0 and 1. If the percentage of labelled cases reaches this value the self-training process is stoped (default: 1) |
verbose |
A boolean indicating the verbosity level of the function. (default: FALSE) |
list |
list of a target class used in the model |
error |
training error |
top |
top proteins |
score |
decision values for top proteins |
Dongmin Jung, Xijin Ge
self.train
data(litG) litG <- igraph.from.graphNEL(litG) sg <- decompose(litG, min.vertices = 50) sg <- sg[[1]] K <- net.kernel(sg) litG.infer.ST <- net.infer.ST(names(V(sg))[1:10], K, top=20)data(litG) litG <- igraph.from.graphNEL(litG) sg <- decompose(litG, min.vertices = 50) sg <- sg[[1]] K <- net.kernel(sg) litG.infer.ST <- net.infer.ST(names(V(sg))[1:10], K, top=20)
This function gives the regularized Laplacian matrix for a graph.
net.kernel(g, decay = 0.5)net.kernel(g, decay = 0.5)
g |
graph |
decay |
decaying constant (default: 0.5) |
the regularized Laplacian matrix
Dongmin Jung, Xijin Ge
laplacian_matrix
# example 1 ## Not run: string.db.9606 <- STRINGdb$new(version = '11', species = 9606, score_threshold = 999) string.db.9606.graph <- string.db.9606$get_graph() K.9606 <- net.kernel(string.db.9606.graph) ## End(Not run) # example 2 data(litG) litG <- igraph.from.graphNEL(litG) sg <- decompose(litG, min.vertices=50) sg <- sg[[1]] K <- net.kernel(sg)# example 1 ## Not run: string.db.9606 <- STRINGdb$new(version = '11', species = 9606, score_threshold = 999) string.db.9606.graph <- string.db.9606$get_graph() K.9606 <- net.kernel(string.db.9606.graph) ## End(Not run) # example 2 data(litG) litG <- igraph.from.graphNEL(litG) sg <- decompose(litG, min.vertices=50) sg <- sg[[1]] K <- net.kernel(sg)
the result from the over-representation analysis
ORA(pathways, gene.id, minSize = 1, maxSize = Inf, p.adjust.methods = NULL)ORA(pathways, gene.id, minSize = 1, maxSize = Inf, p.adjust.methods = NULL)
pathways |
list of gene sets |
gene.id |
set of genes |
minSize |
Minimal size of a gene set |
maxSize |
Maximal size of a gene set |
p.adjust.methods |
a correction method |
ORA result
Dongmin Jung, Xijin Ge
fisher.test
data(examplePathways) data(exampleRanks) geneNames <- names(exampleRanks) set.seed(1) gene.id <- sample(geneNames, 100) ORA(examplePathways, gene.id)data(examplePathways) data(exampleRanks) geneNames <- names(exampleRanks) set.seed(1) gene.id <- sample(geneNames, 100) ORA(examplePathways, gene.id)
For the functional enrichment analysis, we can visualize the result from the over-representation analysis.
ORA.barplot(object, category, size, count, pvalue, top = 10, sort = NULL, decreasing = FALSE, p.adjust.methods = NULL, numChar = NULL, title = NULL, transparency = 0.5, plot = TRUE)ORA.barplot(object, category, size, count, pvalue, top = 10, sort = NULL, decreasing = FALSE, p.adjust.methods = NULL, numChar = NULL, title = NULL, transparency = 0.5, plot = TRUE)
object |
a table with category, size, count and p-value of gene sets |
category |
name of gene sets |
size |
size of gene sets |
count |
count of gene sets |
pvalue |
p-value of gene sets |
top |
the number of top categories (default: 10) |
sort |
a variable used for sorting data |
decreasing |
logical indicating whether ascending or descending order (default: FALSE) |
p.adjust.methods |
a correction method |
numChar |
the maximal number of characters of the name of gene sets |
title |
title for the plot |
transparency |
transparency (default: 0.5) |
plot |
return plot when plot is true, otherwise return table (default: TRUE) |
ORA barplot
Dongmin Jung, Xijin Ge
Yu G, Wang L, Yan G and He Q (2015). "DOSE: an R/Bioconductor package for Disease Ontology Semantic and Enrichment analysis." Bioinformatics, 31(4), pp. 608-609.
p.adjust, ggplot2
data(examplePathways) data(exampleRanks) geneNames <- names(exampleRanks) set.seed(1) gene.id <- sample(geneNames, 100) result.ORA <- ORA(examplePathways, gene.id) ORA.barplot(result.ORA, category = "Category", size = "Size", count = "Count", pvalue = "pvalue", sort = "pvalue")data(examplePathways) data(exampleRanks) geneNames <- names(exampleRanks) set.seed(1) gene.id <- sample(geneNames, 100) result.ORA <- ORA(examplePathways, gene.id) ORA.barplot(result.ORA, category = "Category", size = "Size", count = "Count", pvalue = "pvalue", sort = "pvalue")
This function is designed for human protein-protein interaction from STRING database. Default format is 'hgnc'. The number of proteins is 10 in default. Note that the number of proteins used as a target may be different from the number of proteins in the input since mapping between formats is not always one-to-one in getBM.
ppi.infer.human(target, kernel, top = 10, classifier = net.infer, input = "hgnc_symbol", output = "hgnc_symbol", ...)ppi.infer.human(target, kernel, top = 10, classifier = net.infer, input = "hgnc_symbol", output = "hgnc_symbol", ...)
target |
set of interesting proteins or target class |
kernel |
the regularized Laplacian matrix for a graph |
top |
number of top proteins most closely related to target class (default: 10) |
classifier |
net.infer or net.infer.ST (default: net.infer) |
input |
input format |
output |
output format |
... |
additional parameters for the chosen classifier |
list |
list of a target class used in the model |
error |
training error |
CVerror |
cross validation error, (when cross > 0 in net.infer) |
top |
top proteins |
score |
decision values for top proteins |
Dongmin Jung, Xijin Ge
net.infer, net.infer.ST, getBM
# example 1 string.db.9606 <- STRINGdb$new(version = '11', species = 9606, score_threshold = 999) string.db.9606.graph <- string.db.9606$get_graph() K.9606 <- net.kernel(string.db.9606.graph) rownames(K.9606) <- substring(rownames(K.9606), 6) colnames(K.9606) <- substring(colnames(K.9606), 6) target <- colnames(K.9606)[1:100] infer.human <- ppi.infer.human(target, K.9606, input = "ensembl_peptide_id") ## Not run: # example 2 library(graph) data(apopGraph) target <- nodes(apopGraph) apoptosis.infer <- ppi.infer.human(target, K.9606, 100) # example 3 library(KEGGgraph) library(KEGG.db) pName <- "p53 signaling pathway" pId <- mget(pName, KEGGPATHNAME2ID)[[1]] getKGMLurl(pId, organism = "hsa") p53 <- system.file("extdata/hsa04115.xml", package="KEGGgraph") p53graph <- parseKGML2Graph(p53,expandGenes=TRUE) entrez <- translateKEGGID2GeneID(nodes(p53graph)) httr::set_config(httr::config(ssl_verifypeer = FALSE)) human.ensembl <- useEnsembl(biomart = "ensembl", dataset = "hsapiens_gene_ensembl") target <- getBM(attributes=c('entrezgene', 'hgnc_symbol'), filter = 'entrezgene', values = entrez, mart = human.ensembl)[,2] p53.infer <- ppi.infer.human(target, K.9606, 100) ## End(Not run)# example 1 string.db.9606 <- STRINGdb$new(version = '11', species = 9606, score_threshold = 999) string.db.9606.graph <- string.db.9606$get_graph() K.9606 <- net.kernel(string.db.9606.graph) rownames(K.9606) <- substring(rownames(K.9606), 6) colnames(K.9606) <- substring(colnames(K.9606), 6) target <- colnames(K.9606)[1:100] infer.human <- ppi.infer.human(target, K.9606, input = "ensembl_peptide_id") ## Not run: # example 2 library(graph) data(apopGraph) target <- nodes(apopGraph) apoptosis.infer <- ppi.infer.human(target, K.9606, 100) # example 3 library(KEGGgraph) library(KEGG.db) pName <- "p53 signaling pathway" pId <- mget(pName, KEGGPATHNAME2ID)[[1]] getKGMLurl(pId, organism = "hsa") p53 <- system.file("extdata/hsa04115.xml", package="KEGGgraph") p53graph <- parseKGML2Graph(p53,expandGenes=TRUE) entrez <- translateKEGGID2GeneID(nodes(p53graph)) httr::set_config(httr::config(ssl_verifypeer = FALSE)) human.ensembl <- useEnsembl(biomart = "ensembl", dataset = "hsapiens_gene_ensembl") target <- getBM(attributes=c('entrezgene', 'hgnc_symbol'), filter = 'entrezgene', values = entrez, mart = human.ensembl)[,2] p53.infer <- ppi.infer.human(target, K.9606, 100) ## End(Not run)
This function is designed for mouse protein-protein interaction from STRING database. Default format is 'mgi'. The number of proteins is 10 in default. Note that the number of proteins used as a target may be different from the number of proteins in the input since mapping between formats is not always one-to-one in getBM.
ppi.infer.mouse(target, kernel, top = 10, classifier = net.infer, input = "mgi_symbol", output = "mgi_symbol", ...)ppi.infer.mouse(target, kernel, top = 10, classifier = net.infer, input = "mgi_symbol", output = "mgi_symbol", ...)
target |
set of interesting proteins or target class |
kernel |
the regularized Laplacian matrix for a graph |
top |
number of top proteins most closely related to target class (default: 10) |
classifier |
net.infer or net.infer.ST (default: net.infer) |
input |
input format |
output |
output format |
... |
additional parameters for the chosen classifier |
list |
list of a target class used in the model |
error |
training error |
CVerror |
cross validation error, (when cross > 0 in net.infer) |
top |
top proteins |
score |
decision values for top proteins |
Dongmin Jung, Xijin Ge
net.infer, net.infer.ST, getBM
string.db.10090 <- STRINGdb$new(version = '11', species = 10090, score_threshold = 999) string.db.10090.graph <- string.db.10090$get_graph() K.10090 <- net.kernel(string.db.10090.graph) rownames(K.10090) <- substring(rownames(K.10090), 7) colnames(K.10090) <- substring(colnames(K.10090), 7) target <- colnames(K.10090)[1:100] infer.mouse <- ppi.infer.mouse(target, K.10090, input="ensembl_peptide_id")string.db.10090 <- STRINGdb$new(version = '11', species = 10090, score_threshold = 999) string.db.10090.graph <- string.db.10090$get_graph() K.10090 <- net.kernel(string.db.10090.graph) rownames(K.10090) <- substring(rownames(K.10090), 7) colnames(K.10090) <- substring(colnames(K.10090), 7) target <- colnames(K.10090)[1:100] infer.mouse <- ppi.infer.mouse(target, K.10090, input="ensembl_peptide_id")
This function can be used for classification of semi-supervised data by using the kernel support vector machine.
self.train.kernel(K, y, type = 'response', C = 1, cache = 40, tol = 0.001, shrinking = TRUE, thrConf = 0.9, maxIts = 10, percFull = 1, verbose = FALSE)self.train.kernel(K, y, type = 'response', C = 1, cache = 40, tol = 0.001, shrinking = TRUE, thrConf = 0.9, maxIts = 10, percFull = 1, verbose = FALSE)
K |
kernel matrix |
y |
lable vector |
type |
one of response, probabilities ,votes, decision indicating the type of output (default: response) |
C |
cost of constraints violation for SVM (default: 1) |
cache |
cache memory in MB for SVM (default: 40) |
tol |
tolerance of termination criterion for SVM (default: 0.001) |
shrinking |
option whether to use the shrinking-heuristics for OCSVM (default: TRUE) |
thrConf |
A number between 0 and 1, indicating the required classification confidence for an unlabelled case to be added to the labelled data set with the label predicted predicted by the classification algorithm (default: 0.9) |
maxIts |
The maximum number of iterations of the self-training process (default: 10) |
percFull |
A number between 0 and 1. If the percentage of labelled cases reaches this value the self-training process is stoped (default: 1) |
verbose |
A boolean indicating the verbosity level of the function (default: FALSE) |
prediction from the SVM
Dongmin Jung, Xijin Ge
Torgo, L. (2016) Data Mining using R: learning with case studies, second edition, Chapman & Hall/CRC.
data(litG) litG <- igraph.from.graphNEL(litG) sg <- decompose(litG, min.vertices = 50) sg <- sg[[1]] K <- net.kernel(sg) y <- rep(NA, length(V(sg))) y[1:10] <- 1 y[11:20] <- 0 y <- factor(y) self.train.kernel(K, y)data(litG) litG <- igraph.from.graphNEL(litG) sg <- decompose(litG, min.vertices = 50) sg <- sg[[1]] K <- net.kernel(sg) y <- rep(NA, length(V(sg))) y[1:10] <- 1 y[11:20] <- 0 y <- factor(y) self.train.kernel(K, y)