Type: Package
Title: A Copy Number and Expression-Based Classifier for Breast Tumours
Version: 2.0.2
Date: 2024-07-16
Maintainer: Oscar M Rueda <Oscar.Rueda@mrc-bsu.cam.ac.uk>
Description: Implementation of the classifier described in the paper Ali HR et al (2014) <doi:10.1186/s13059-014-0431-1>. It uses copy number and/or expression form breast cancer data, trains a Tibshirani's 'pamr' classifier with the features available and predicts the iC10 group.
License: GPL-3
Imports: pamr, impute, iC10TrainingData
Packaged: 2024-07-19 06:32:13 UTC; oscar
NeedsCompilation: no
Repository: CRAN
Date/Publication: 2024-07-19 09:00:26 UTC
Author: Oscar M Rueda ORCID iD [aut, cre]

A Copy Number and Expression-Based Classifier for Breast Tumours

Description

Implementation of the classifier described in the paper Ali HR et al (2014) <doi:10.1186/s13059-014-0431-1>. It uses copy number and/or expression form breast cancer data, trains a Tibshirani's 'pamr' classifier with the features available and predicts the iC10 group.

Details

The DESCRIPTION file:

Package: iC10
Type: Package
Title: A Copy Number and Expression-Based Classifier for Breast Tumours
Version: 2.0.2
Date: 2024-07-16
Authors@R: person("Oscar M", "Rueda", , "Oscar.Rueda@mrc-bsu.cam.ac.uk", role = c("aut", "cre"), comment = c(ORCID = "0000-0003-0008-4884"))
Maintainer: Oscar M Rueda <Oscar.Rueda@mrc-bsu.cam.ac.uk>
Description: Implementation of the classifier described in the paper Ali HR et al (2014) <doi:10.1186/s13059-014-0431-1>. It uses copy number and/or expression form breast cancer data, trains a Tibshirani's 'pamr' classifier with the features available and predicts the iC10 group.
License: GPL-3
Imports: pamr, impute, iC10TrainingData
Packaged: 2015-09-23 02:37:41 UTC; rueda01
NeedsCompilation: no
Repository: CRAN
Date/Publication: 2015-09-23 08:05:08
Author: Oscar M Rueda [aut, cre] (<https://orcid.org/0000-0003-0008-4884>)

Index of help topics: 

compare                 Compare results of the iC10 classifier
getCNfeatures           Internal function for matching copy number
                        features.
getExpfeatures          Internal function for matching expression
                        features.
goodnessOfFit           Goodness of fit results of the iC10 classifier
iC10                    A copy number and expression-based classfier
                        for breast cancers
iC10-package            A Copy Number and Expression-Based Classifier
                        for Breast Tumours
matchFeatures           Matching features from the classifier to the
                        test data.
normalizeFeatures       Normalization of expression features
plot.iC10               Plot results of the iC10 classifier
print.iC10              Print results of the iC10 classifier
summary.iC10            Summary results of the iC10 classifier

iC10 implements the classifier described in the paper 'Genome-driven integrated classification of breast cancer validated in over 7,500 samples' (Ali HR et al., Genome Biology 2014). It uses copy number and/or expression form breast cancer data, trains a pamr classifier (Tibshirani et al.) with the features available and predicts the iC10 group.

Author(s)

NA

Maintainer: Oscar M Rueda <Oscar.Rueda@mrc-bsu.cam.ac.uk>

References

Ali HR et al. Genome-driven integrated classification of breast cancer validated in over 7,500 samples. Genome Biology 2014; 15:431. Curtis et al. The genomic and transcriptomic architecture of 2,000 breast tumours reveals novel subgroups. Nature 2012; 486:346-352. Tibshirani et al. Diagnosis of multiple cancer types by shrunken centroids of gene expression. PNAS 2002; 99(10):6567-6572.

Examples

require(iC10TrainingData)
data(train.CN)
data(train.Exp)
features <- matchFeatures(Exp=train.Exp, Exp.by.feat="probe")
features <- normalizeFeatures(features, "scale")
res <- iC10(features)
summary(res)
goodnessOfFit(res, newdata=features)
compare(res, iC10=1:2, newdata=features)
compare(res, iC10=2:4, newdata=features)

Compare results of the iC10 classifier

Description

This function plots the centroids of the training set versus the average profiles of the new data classified in each group.

Usage

compare(obj, iC10=1:10, newdata, name.test="Test",...)
## S3 method for class 'iC10'
compare(obj, iC10=1:10, newdata, name.test="Test",...)

Arguments

obj

An object of class iC10, a result of a call to iC10()

iC10

Groups to plot

newdata

Set of features of the new data to compare. They must be the same samples classified and contained in x. A result of a call to matchFeatures() or normalizeFeatures()

name.test

Name of the new data set to appear in the text of the plot

...

Additional arguments passed to plot()

Value

A plot is returned with two plots per groups requested.

Author(s)

Oscar M. Rueda

References

Ali HR et al. Genome-driven integrated classification of breast cancer validated in over 7,500 samples. Genome Biology 2014; 15:431. Curtis et al. The genomic and transcriptomic architecture of 2,000 breast tumours reveals novel subgroups. Nature 2012; 486:346-352.

See Also

iC10, plot.iC10, matchFeatures, normalizeFeatures

Examples

require(iC10TrainingData)
data(train.CN)
data(train.Exp)
features <- matchFeatures(Exp=train.Exp, Exp.by.feat="probe")
features <- normalizeFeatures(features, "scale")
res <- iC10(features)
compare(res, 1:3, newdata=features)

Internal function for matching copy number features.

Description

This function should not be called directly

Usage

getCNfeatures(CN, Probes, Map, by.feat, ref, Synonyms)

Arguments

CN

CN features matrix

Probes

Vector with the probes to match

Map

data.frame with the genomic description of the features to match

by.feat

"probe" or "gene", indicating if match should be done by probe position or gene name.

ref

hg18 or hg19 (only relevant if matching is done by probe position).

Synonyms

data.frame with available synonym gene names to match (only relevant if matching is done by gene name).

Value

A matrix with the copy number features

Author(s)

Oscar M Rueda

Internal function for matching expression features.

Description

Internal function for matching expression features.

Usage

getExpfeatures(Exp, Probes, Synonyms, by.feat)

Arguments

Exp

Matrix of expression features

Probes

Vector of probes to match

Synonyms

vector of synonyms fo gene names

by.feat

either "probe" or "gene"

Value

A matrix with the Probes in Exp.

Note

This function is not supposed to be called directly. use matchFeatures instead.

Author(s)

Oscar M Rueda

References

Curtis et al. The genomic and transcriptomic architecture of 2,000 breast tumours reveals novel subgroups. Nature 2012; 486:346-352.

See Also

matchFeatures

Goodness of fit results of the iC10 classifier

Description

Goodness of fit results of the iC10 classifier: this function computes correlations between the signatures of the training dataset and the classified features.

Usage

goodnessOfFit(obj, iC10=1:10, newdata=NULL,...)
## S3 method for class 'iC10'
goodnessOfFit(obj, iC10=1:10, newdata=NULL,...)

Arguments

obj

An object of iC10 class.

iC10

Groups to compute goodness of fit.

newdata

The feature data to compute the goodness of fit. Must be the samples classified in obj. It can be a call to matchFeatures or normalizeFeatures. If NULL, obj$fitted is used.

...

Additional arguments passed to cor (like method; Default is pearson)

Value

It prints the correlation for each iC10.

Author(s)

Oscar M Rueda

References

Ali HR et al. Genome-driven integrated classification of breast cancer validated in over 7,500 samples. Genome Biology 2014; 15:431. Curtis et al. The genomic and transcriptomic architecture of 2,000 breast tumours reveals novel subgroups. Nature 2012; 486:346-352.

See Also

iC10

Examples

require(iC10TrainingData)
data(train.CN)
data(train.Exp)
features <- matchFeatures(Exp=train.Exp, Exp.by.feat="probe")
features <- normalizeFeatures(features, "scale")
res <- iC10(features)
goodnessOfFit(res, newdata=features)

A copy number and expression-based classfier for breast cancers

Description

iC10 implements the classifier described in the paper 'Genome-driven integrated classification of breast cancer validated in over 7,500 samples' (Ali HR et al., Genome Biology 2014). It uses copy number and/or expression form breast cancer data, trains a pamr classifier (Tibshirani et al.) with the features available and predicts the iC10 group.

Usage

iC10(x, seed=25435)

Arguments

x

An object with class iC10features: A list with elements 'train.CN', 'train.Exp', 'train.iC10', 'CN', 'Exp', 'map.cn', 'map.exp'

seed

seed to initialize random number generator. It is passed to set.seed(). See details.

Details

This function trains a pamr classifier and predicts the set of samples. The shrinkage parameter is obtained with crossvalidation, therefore different runs can give different results (unless a seed is specified).

Value

An object of class iC10. A list with the following elements:

class

Prediction classes for the samples

posterior

Probablitites for each sample to belong to each of the 10 groups

centroids

Shrunken Centroids for each of the 10 groups.

fitted

Normalized features for the samples classified.

map.cn

Annotation data for the copy number features

map.exp

Annotation data for the expression features

Author(s)

Oscar M. Rueda

References

Ali HR et al. Genome-driven integrated classification of breast cancer validated in over 7,500 samples. Genome Biology 2014; 15:431. Curtis et al. The genomic and transcriptomic architecture of 2,000 breast tumours reveals novel subgroups. Nature 2012; 486:346-352. Tibshirani et al. Diagnosis of multiple cancer types by shrunken centroids of gene expression. PNAS 2002; 99(10):6567-6572.

See Also

See pamr.train, pamr.cv and pamr.predict in package pamr.

Examples

require(iC10TrainingData)
data(train.CN)
data(train.Exp)
features <- matchFeatures(Exp=train.Exp, Exp.by.feat="probe")
features <- normalizeFeatures(features, "scale")
res <- iC10(features)

Matching features from the classifier to the test data.

Description

This function matches available copy number and/or expression data features to the training signatures; using either genomic position or HUGO gene name for copy number features and either Illumina probe names or HUGO gene name for expression features.

Usage

matchFeatures(CN = NULL, Exp = NULL,
CN.by.feat = c("gene", "probe"),
Exp.by.feat = c("gene", "probe"),
ref="hg19")

Arguments

CN

Data must be log2 copy number ratios. Two formats are allowed: - a matrix where each row represents a gene and each column a sample. In this case CN.by.feat must be "gene" and the rownames must be the hgnc gene names. - a data.frame with segmented data. The following columns must exist: 'ID' for the sample name, 'chromosome_name' for the chromosome (must be numeric), 'loc.start' for the start position of the region, 'loc.end' for the end position of the region, 'seg.mean' for the log2ratio of the segment. If NULL, copy number is not used in the classifier.

Exp

Matrix with the expression data to classify. Each row must be a gene or an Illumina probe, and each column must correspond to a sample. Rownames must be either Illumina probes, in which case Exp.by.feat must be "probe"; or hgnc gene names, in which case Exp.by.feat must be "gene". If NULL, expression is not used in the classifier.

CN.by.feat

Either "probe" or "gene", Default is "probe".

Exp.by.feat

Either "probe" or "gene", Default is "gene".

ref

Either "hg18", "hg19" or "hg38". It is used to match the copy number probes if CN.by.feat is "probe"

Details

One of CN or Exp must be not NULL. If matching is done by gene, hgnc gene name is used to match the rownames of the features. A list of synonym gene names is used (see Map.All). For copy number features matched by probe, the maximum log ratio in absolute value inside the limits of the feature is used. If there is no copy number in that region, the value of the probe before it is used.

Value

A list with the following elements is returned:

CN

copy number data to classify

train.CN

copy number training data

Exp

expression data to classify

train.Exp

expression training data

train.iC10

iC10 assignments for the training data

map.cn

annotation data for the copy number features

map.exp

annotation data for the expression features

Note

Note that the training set will be different, depending on the features matched. Genomic annotation for the training dataset has been obtained from Mark Dunning's lluminaHumanv3.db package.

Author(s)

Oscar M Rueda

References

Ali HR et al. Genome-driven integrated classification of breast cancer validated in over 7,500 samples. Genome Biology 2014; 15:431. Curtis et al. The genomic and transcriptomic architecture of 2,000 breast tumours reveals novel subgroups. Nature 2012; 486:346-352.

See Also

normalizeFeatures, iC10

Examples

require(iC10TrainingData)
data(train.CN)
data(train.Exp)
features <- matchFeatures(Exp=train.Exp,Exp.by.feat="probe", ref="hg18")
str(features)

Normalization of expression features

Description

Normalization of expression features. Several methods available in the package CONOR can be used.

Usage

normalizeFeatures(x, method=c("none", "scale"))

Arguments

x

An object result of a call to matchFeatures

method

Several methods are available: "none": No normalization is done "scale": Each expression feature is scaled to have zero mean and standard deviation 1

Details

No further normalization is needed on the copy number, as log2 ratios are comparable between platforms.

Value

A list of the same format as matchFeatures, but with train.Exp anfd Exp normalized.

Note

As CONOR package is no longer maintained, the methods are not available temporarily. We will include more normalization methods in the next version of this package.

Author(s)

Oscar M Rueda

References

Ali HR et al. Genome-driven integrated classification of breast cancer validated in over 7,500 samples. Genome Biology 2014; 15:431. Curtis et al. The genomic and transcriptomic architecture of 2,000 breast tumours reveals novel subgroups. Nature 2012; 486:346-352.

Examples

require(iC10TrainingData)
data(train.CN)
data(train.Exp)
features <- matchFeatures(Exp=train.Exp,
Exp.by.feat="probe", ref="hg18")
features <- normalizeFeatures(features, "scale")

Plot results of the iC10 classifier

Description

Plot results of the iC10 classifier, in two different formats: either the signatures of the training set or the signatures of the new data classified.

Usage

## S3 method for class 'iC10'
plot(x, sample.name=1, newdata = NULL,...)

Arguments

x

An object of iC10 class:

sample.name

Number of sample to plot (if newdata is NULL). It can be either a number or a character with the sample name.

newdata

An object result to call to matchFeatures or normalizeFeatures containing the features of the samples to plot.

...

Additional arguments passed to plot.

Details

Two types of plots can be produced. If newdata is NULL, a panel 6x2 is drawn with the 10 profiles of the signatures of the training set and the profile of the features of sample.name and the distribution of the probabilities of classification to each iC10 for that sample. If newdata is not nutll, a panel 6x2 (with the 11th panel empty) is drawn with the 10 profiles of newdata samples and their distribution into the clusters. The features are sorted by type: copy number (if available) are drawn in grey, and then expression, each of them are sorted by genomic position.

Value

A 6x2 plot is produced.

Author(s)

Oscar M Rueda

References

Ali HR et al. Genome-driven integrated classification of breast cancer validated in over 7,500 samples. Genome Biology 2014; 15:431. Curtis et al. The genomic and transcriptomic architecture of 2,000 breast tumours reveals novel subgroups. Nature 2012; 486:346-352.

See Also

iC10

Examples

require(iC10TrainingData)
data(train.CN)
data(train.Exp)
features <- matchFeatures(Exp=train.Exp, Exp.by.feat="probe")
features <- normalizeFeatures(features, "scale")
res <- iC10(features)
plot(res, sample.name=10)
plot(res, newdata=features)

Print results of the iC10 classifier

Description

Print results of the iC10 classifier

Usage

## S3 method for class 'iC10'
print(x, ...)

Arguments

x

An object of iC10 class:

...

Additional arguments passed to print.

Value

It returns a call to str.

Author(s)

Oscar M Rueda

References

Ali HR et al. Genome-driven integrated classification of breast cancer validated in over 7,500 samples. Genome Biology 2014; 15:431. Curtis et al. The genomic and transcriptomic architecture of 2,000 breast tumours reveals novel subgroups. Nature 2012; 486:346-352.

See Also

iC10

Examples

require(iC10TrainingData)
data(train.CN)
data(train.Exp)
features <- matchFeatures(Exp=train.Exp, Exp.by.feat="probe")
features <- normalizeFeatures(features, "scale")
res <- iC10(features)
res

Summary results of the iC10 classifier

Description

Summary results of the iC10 classifier: shows the distribution of samples classified into each iC10 group and a summary of the maximum posterior probablity for each sample. Small values pinpoint samples with no clear group assigned.

Usage

## S3 method for class 'iC10'
summary(object, ...)

Arguments

object

An object of iC10 class.

...

Additional arguments passed to summary.

Value

The function prints a table of the classification ad a summary of the maximum posterior probability for each sample.

Author(s)

Oscar M Rueda

References

Ali HR et al. Genome-driven integrated classification of breast cancer validated in over 7,500 samples. Genome Biology 2014; 15:431. Curtis et al. The genomic and transcriptomic architecture of 2,000 breast tumours reveals novel subgroups. Nature 2012; 486:346-352. Tibshirani et al. Diagnosis of multiple cancer types by shrunken centroids of gene expression. PNAS 2002; 99(10):6567-6572.

See Also

See iC10 and pamr.train, pamr.cv and pamr.predict in package pamr.

Examples

require(iC10TrainingData)
data(train.CN)
data(train.Exp)
features <- matchFeatures(Exp=train.Exp,
Exp.by.feat="probe", ref="hg18")
features <- normalizeFeatures(features, "scale")
res <- iC10(features)
summary(res)