| Title: | Poor prognosis colon cancer is defined by a molecularly distinct subtype and precursor lesion |
|---|---|
| Description: | This package reproduces the main pipeline to analyze the AMC-AJCCII-90 microarray data set in De Sousa et al. accepted by Nature Medicine in 2013. |
| Authors: | Xin Wang <[email protected]> |
| Maintainer: | Xin Wang <[email protected]> |
| License: | Artistic-2.0 |
| Version: | 1.48.0 |
| Built: | 2026-10-10 11:07:45 UTC |
| Source: | https://github.com/bioc/DeSousa2013 |
This function employs PAM to build a gene expression based classifier.
buildClassifier(sigMat, clus.f, nfold=10, nboot=100) figPAMCV(err)buildClassifier(sigMat, clus.f, nfold=10, nboot=100) figPAMCV(err)
sigMat |
a matrix of median centered expression values of selected genes (by function
|
clus.f |
a numeric vector of cluster labels for selected cancer samples. |
nfold |
an integer value specifying the fold of cross validation. |
nboot |
an integer value specifying the number of bootstraps. |
err |
a matrix of cross validation error rates for different shrinkage thresholds. |
The expression data of the retained most predictive genes were trained by PAM to build a robust classifier. To select the optimal threshold for centroid shrinkage, we performed 10-fold cross-validation over a range of shirnkage thresholds for 1000 iterations, and selected the one yielding a good performance (error rate < 2%) with the least number of genes.
This function will return a list including signature (a character vector
of signature genes), pam.rslt (an list of training results returned by
pamr.train), thresh (the selected shrinkage
threshold), err (a matrix of cross validation error rates for different
shrinkage thresholds), cents (a numeric matrix of PAM centroids for three
subtypes).
Xin Wang [email protected]
De Sousa E Melo, F. and Wang, X. and Jansen, M. et al. Poor prognosis colon cancer is defined by a molecularly distinct subtype and precursor lesion. accepted
Tibshirani, Robert and Hastie, Trevor and Narasimhan, Balasubramanian and Chu, Gilbert (2002). Diagnosis of multiple cancer types by shrunken centroids of gene expression. PNAS, 99(10), 6567-6572.
filterSamples, findDiffGenes, filterDiffGenes
data(diffGenes.f, package="DeSousa2013") data(silh, package="DeSousa2013") sigMat <- sdat.f[diffGenes.f, names(clus.f)] ##set a small number of bootstraps, only for testing the function classifier <- buildClassifier(sigMat, clus.f, nboot=10) data(classifier, package="DeSousa2013") figPAMCV(err)data(diffGenes.f, package="DeSousa2013") data(silh, package="DeSousa2013") sigMat <- sdat.f[diffGenes.f, names(clus.f)] ##set a small number of bootstraps, only for testing the function classifier <- buildClassifier(sigMat, clus.f, nboot=10) data(classifier, package="DeSousa2013") figPAMCV(err)
Compute Gap statistics to identify the optimal number of subtypes
compGapStats(ge.CRC, ntops=c(2, 4, 8, 12, 16, 20)*1000, K.max=6, nboot=100) figGAP(gapsmat, gapsSE)compGapStats(ge.CRC, ntops=c(2, 4, 8, 12, 16, 20)*1000, K.max=6, nboot=100) figGAP(gapsmat, gapsSE)
ge.CRC |
a numeric matrix of expression data of genes expressed in at least one sample. |
ntops |
an integer vector of top variable genes, measured by MAD (median absolute deviation). |
K.max |
an integer value specifying the maximal number of clusters to compute GAP statistics. |
nboot |
an integer value specifying the number of bootstraps, which is an argument |
gapsmat |
a numeric matrix of GAP statistics. |
gapsSE |
standard errors of means of the GAP statistics. |
GAP statistic is a popular method to estimate the number of clusters in a set
of data by comparing the change in observed and expected within-cluster
dispersion. To identify the optimal number of clusters, GAP statistic can be
computed for k=1 to K.max with nboot bootstraps for ntops
top variable genes in the AMC data set.
The function figGAP is designed to visualize GAP curves.
This function will return a list including gapsmat (a numeric matrix of
GAP statistics) and gapsSE (standard errors of means of the GAP
statistics).
Xin Wang [email protected]
De Sousa E Melo, F. and Wang, X. and Jansen, M. et al. Poor prognosis colon cancer is defined by a molecularly distinct subtype and precursor lesion. accepted
Tibshirani, Robert andWalther, Guenther and Hastie, Trevor (2001). Estimating the number of clusters in a data set via the gap statistic. Journal of the Royal Statistical Society: Series B (Statistical Methodology), 63(2), 411-423.
data(ge.CRC, package="DeSousa2013") ge.CRC <- ge.all[selPbs, ] gaps <- compGapStats(ge.CRC, ntops=c(2, 4)*1000, K.max=6, nboot=10) figGAP(gaps$gapsmat, gaps$gapsSE)data(ge.CRC, package="DeSousa2013") ge.CRC <- ge.all[selPbs, ] gaps <- compGapStats(ge.CRC, ntops=c(2, 4)*1000, K.max=6, nboot=10) figGAP(gaps$gapsmat, gaps$gapsSE)
This function performs consensus clustering for the AMC-AJCCII-90 data set.
conClust(sdat, maxK=12, reps=1000, savepath = ".")conClust(sdat, maxK=12, reps=1000, savepath = ".")
sdat |
a matrix of median centered expression values of top variable probesets. |
maxK |
an integer value specifying the maximal number of clusters to perform consensus clustering. |
reps |
an integer value specifying resampling times. |
savepath |
the path to a directory where figures will be saved. |
Using the most variable probesets (MAD>0.5), we performed hierarchical clustering with agglomerative average linkage to cluster these samples. Consensus clustering was employed, with 1000 iterations and 0.98 subsampling ratio, to assess the clustering stability.
This function will return clus, which is a numeric vector of cluster
labels named by colon cancer samples.
Xin Wang [email protected]
De Sousa E Melo, F. and Wang, X. and Jansen, M. et al. Poor prognosis colon cancer is defined by a molecularly distinct subtype and precursor lesion. accepted
Monti, Stefano and Tamayo, Pablo and Mesirov, Jill and Golub, Todd (2003). Consensus clustering: a resampling-based method for class discovery and visualization of gene expression microarray data. Machine learning, 52(1), 91-118.
library(ConsensusClusterPlus) data(dat, package="DeSousa2013") ##set a small number of bootstraps only for testing the function clus <- conClust(sdat, maxK=6, reps=10, savepath=".")library(ConsensusClusterPlus) data(dat, package="DeSousa2013") ##set a small number of bootstraps only for testing the function clus <- conClust(sdat, maxK=6, reps=10, savepath=".")
This function reproduces the main pipeline to analyze the AMC-AJCCII-90 microarray data set in De Sousa et al. accepted by Nature Medicine in 2013.
CRCPipeLine(celpath=".", AMC_sample_head, AMC_CRC_clinical, preprocess=FALSE, gap.ntops = c(2, 4, 8, 12, 16, 20)*1000, gap.K.max = 6, gap.nboot = 100, MADth=0.5, conClust.maxK=12, conClust.reps=1000, diffG.pvalth=0.01, diffG.aucth=0.9, savepath=".")CRCPipeLine(celpath=".", AMC_sample_head, AMC_CRC_clinical, preprocess=FALSE, gap.ntops = c(2, 4, 8, 12, 16, 20)*1000, gap.K.max = 6, gap.nboot = 100, MADth=0.5, conClust.maxK=12, conClust.reps=1000, diffG.pvalth=0.01, diffG.aucth=0.9, savepath=".")
celpath |
the path to a directory where .CEL files of the AMC-AJCCII-90 set are placed. |
AMC_sample_head |
a data frame include clinical information and mapping information between
microarray ids and sample ids (details in |
AMC_CRC_clinical |
a data frame including include clinical information (details in |
preprocess |
a logical value specifying whether preprocessing of microarrays should be
performed. If FALSE (default), preprocessed expression data |
gap.ntops |
an integer vector of top variable genes, measured by MAD (median absolute deviation). |
gap.K.max |
an integer value specifying the maximal number of clusters to compute GAP statistics. |
gap.nboot |
an integer value specifying the number of bootstraps, which is an argument |
MADth |
an numeric value specifying the cutoff of MAD (median absolute deviation). |
conClust.maxK |
an integer value specifying the maximal number of clusters to perform consensus clustering. |
conClust.reps |
an integer value specifying resampling times. |
diffG.pvalth |
a numeric value specifying the fdr cutoff to select differential genes between subtypes by SAM. |
diffG.aucth |
a numeric value specifying the AUC cutoff. |
savepath |
the path to a directory where figures will be saved. |
The function wraps up the preprocessing, feature selection, classification and subtype clinical characterization steps altogether in a signal function. It reproduces the main results and figures of De Sousa 2013, Nature Medicine.
This function will save figure of gap statistics, Silhouette information, PAM
cross-validation error rates, classification heatmap as well as Kaplan Meier
plot to savepath.
Xin Wang [email protected]
De Sousa E Melo, F. and Wang, X. and Jansen, M. et al. Poor prognosis colon cancer is defined by a molecularly distinct subtype and precursor lesion. accepted
geneExpPre, compGapStats, selTopVarGenes,
pbs2unigenes, conClust, filterSamples,
findDiffGenes, filterDiffGenes, buildClassifier,
pamClassify, progAMC
## Not run: ##This function may take a long time (hours) to finish. ##Please use R-2.15 and corresponding frma package to run the function. ##Please contact the author if there is any confusion data(AMC, package="DeSousa2013") CRCPipeLine(celpath=".", AMC_sample_head, AMC_CRC_clinical, savepath=".") ## End(Not run)## Not run: ##This function may take a long time (hours) to finish. ##Please use R-2.15 and corresponding frma package to run the function. ##Please contact the author if there is any confusion data(AMC, package="DeSousa2013") CRCPipeLine(celpath=".", AMC_sample_head, AMC_CRC_clinical, savepath=".") ## End(Not run)
See ‘details’ for the description of each data files included in this package.
##see example for details##see example for details
AMC: AMC_CRC_clinical and AMC_sample_head include clinical information and
mapping information between microarray ids and sample ids, respectively.
ge.CRC: ge.all is a numeric matrix including expression levels of all probesets
of the 90 colon cancer samples. selPbs is a character vector of probeset ids
that are present in any sample.
gaps: gapsmat is a numeric matrix of GAP statistics. gapsSE is a numeric matrix
of standard errors of means of the GAP statistics.
dat: sdat is a matrix of median centered expression values of top variable
probesets.
uniGenes: uniGenes is a character vector of unique gene symbols named by
probesets.
conClust: clus is a numeric vector of cluster labels named by colon cancer
samples.
silh: sdat.f is a matrix of median centered expression values of top variable
genes for selected samples. clus.f is a numeric vector of cluster labels for
selected cancer samples. silh is an object of class silhouette.
diffGenes: diffGenes is a character vector of differential genes.
diffGenes.f: diffGenes.f is a character vector of most predictive genes.
classifier: signature is a character vector of signature genes. pam.rslt is
an list of training results returned by pamr.train.
thresh is the selected shrinkage threshold. err is a matrix of cross validation
error rates for different shrinkage thresholds. cents is a numeric matrix of
PAM centroids for three subtypes.
predAMC: sdat.sig is a numeric matrix of expression values of signature genes
for the 90 samples. pred is a numeric matrix of posterior probabilities for
samples to be classified to subtypes. clu.pred is a numeric vector of
classification labels named by colon cancer samples. nam.ord is a character
vector of samples ordered by their classification probabilies for visualization.
gclu.f is the result of hierarchical clustering on the expression of signature
genes for visualization.
survival: surv is the result of survfit containing survival
curves of the AMC data set. survstats is the result of survdiff.
data4surv is the data to perform survival analysis.
Xin Wang [email protected]
De Sousa E Melo, F. and Wang, X. and Jansen, M. et al. Poor prognosis colon cancer is defined by a molecularly distinct subtype and precursor lesion. accepted
data(AMC, package="DeSousa2013")data(AMC, package="DeSousa2013")
This function evaluates AUC (area under ROC curve) to select the most predictive genes from differential genes.
filterDiffGenes(sdat.f, clus.f, diffGenes, aucth=0.9)filterDiffGenes(sdat.f, clus.f, diffGenes, aucth=0.9)
sdat.f |
a matrix of median centered expression values of top variable genes for selected samples. |
clus.f |
a numeric vector of cluster labels for selected cancer samples. |
diffGenes |
a character vector of differential genes. |
aucth |
a numeric value specifying the AUC cutoff. |
After obtaining differential genes between subtypes, we calculated AUC
(area under ROC curve, using package ROCR) to assess each gene's ability
to separate one subtype from the others.
This function will return diffGenes.f, which is a character vector of
most predictive genes.
Xin Wang [email protected]
De Sousa E Melo, F. and Wang, X. and Jansen, M. et al. Poor prognosis colon cancer is defined by a molecularly distinct subtype and precursor lesion. accepted
Sing, T., Sander, O., Beerenwinkel, N., Lengauer, T. (2005). ROCR: visualizing classifier performance in R. Bioinformatics 21(20):3940-3941.
data(uniGenes) data(silh) data(diffGenes) ##select randomly part of the whole differential genes only for testing the function diffGenes <- diffGenes[sample(1:length(diffGenes), 100)] diffGenes.f <- filterDiffGenes(sdat.f, clus.f, diffGenes, aucth=0.9)data(uniGenes) data(silh) data(diffGenes) ##select randomly part of the whole differential genes only for testing the function diffGenes <- diffGenes[sample(1:length(diffGenes), 100)] diffGenes.f <- filterDiffGenes(sdat.f, clus.f, diffGenes, aucth=0.9)
This function computes Silhouette widths for the 90 colon cancer samples.
filterSamples(sdat, uniGenes, clus) figSilh(silh)filterSamples(sdat, uniGenes, clus) figSilh(silh)
sdat |
a matrix of median centered expression values of top variable probesets. |
uniGenes |
a character vector of unique gene symbols named by probesets. |
clus |
a numeric vector of cluster labels named by colon cancer samples. |
silh |
an object of class |
Silhouette width was computed to identify the most representative samples within each cluster. Samples with positive silhouette width were retained to build the PAM classifier.
This function will return a list including sdat.f (a matrix of median
centered expression values of top variable genes for selected samples),
clus.f (a numeric vector of cluster labels for selected cancer samples)
and silh (an object of class silhouette).
Xin Wang [email protected]
De Sousa E Melo, F. and Wang, X. and Jansen, M. et al. Poor prognosis colon cancer is defined by a molecularly distinct subtype and precursor lesion. accepted
Rousseeuw, Peter J (1987). Silhouettes: a graphical aid to the interpretation and validation of cluster analysis Journal of computational and applied mathematics, 20, 53-65.
data(uniGenes, package="DeSousa2013") data(dat, package="DeSousa2013") data(conClust, package="DeSousa2013") samp.f <- filterSamples(sdat, uniGenes, clus) figSilh(samp.f$silh)data(uniGenes, package="DeSousa2013") data(dat, package="DeSousa2013") data(conClust, package="DeSousa2013") samp.f <- filterSamples(sdat, uniGenes, clus) figSilh(samp.f$silh)
This function employs SAM to search for differential genes between subtypes
findDiffGenes(sdat.f, clus.f, pvalth=0.01)findDiffGenes(sdat.f, clus.f, pvalth=0.01)
sdat.f |
a matrix of median centered expression values of top variable genes for selected samples. |
clus.f |
a numeric vector of cluster labels for selected cancer samples. |
pvalth |
a numeric value specifying the fdr cutoff to select differential genes between subtypes by SAM. |
In this function, we find differential genes between each two of three subtypes and take the unique genes of combined differential genes altogether.
This function will return diffGenes, which is a character vector of
differential genes.
Xin Wang [email protected]
De Sousa E Melo, F. and Wang, X. and Jansen, M. et al. Poor prognosis colon cancer is defined by a molecularly distinct subtype and precursor lesion. accepted
Tusher, Virginia Goss and Tibshirani, Robert and Chu, Gilbert (2001). Significance analysis of microarrays applied to the ionizing radiation response. PNAS, 98(9), 5116-5121.
data(uniGenes, package="DeSousa2013") data(conClust, package="DeSousa2013") data(silh, package="DeSousa2013") ##select randomly part of the whole data set only for testing the function diffGenes <- findDiffGenes(sdat.f[sample(1:nrow(sdat.f), 500), ], clus.f, pvalth=0.01)data(uniGenes, package="DeSousa2013") data(conClust, package="DeSousa2013") data(silh, package="DeSousa2013") ##select randomly part of the whole data set only for testing the function diffGenes <- findDiffGenes(sdat.f[sample(1:nrow(sdat.f), 500), ], clus.f, pvalth=0.01)
Preprocessing of the AMC-AJCCII-90 microarray data
geneExpPre(celpath, AMC_sample_head)geneExpPre(celpath, AMC_sample_head)
celpath |
the path to a directory where .CEL files of the AMC-AJCCII-90 set are placed. |
AMC_sample_head |
a data frame include clinical information and mapping information between
microarray ids and sample ids (details in |
This function reproduces the preprocessing of the AMC-AJCCII-90 colon cancer
microarrays, together with 13 adenomas and 6 normal samples. The microarrays of
the 90 cancer samples and normal samples were generated in one batch, and the
adenoma samples in a different batch. These two batches were first
normalized separately by frma, and then corrected for batch
effect using ComBat. The barcode
function is used to call present probesets in the two batches,
respectively. Finally, probesets present in any one of the cancer, adenoma or
normal samples are selected for the following analysis.
This function will return a list including ge.all (a numeric matrix
including expression levels of all probesets of the 90 colon cancer samples)
and selPbs (a character vector of probeset ids that are present in any
sample).
Xin Wang [email protected]
De Sousa E Melo, F. and Wang, X. and Jansen, M. et al. Poor prognosis colon cancer is defined by a molecularly distinct subtype and precursor lesion. accepted
McCall, Matthew N and Bolstad, Benjamin M and Irizarry, Rafael A (2010). Frozen robust multiarray analysis (fRMA). Biostatistics, 11(2), 242-253.
McCall, Matthew N and Uppal, Karan and Jaffee, Harris A and Zilliox, Michael J and Irizarry, Rafael A (2011). The Gene Expression Barcode: leveraging public data repositories to begin cataloging the human and murine transcriptomes. Nucleic acids research, 39(suppl 1), D1011-D1015.
Johnson, WE, Rabinovic, A, and Li, C (2007). Adjusting batch effects in microarray expression data using Empirical Bayes methods. Biostatistics 8(1):118-127.
## Not run: ##Please make sure that all colon cancer, adenomas as well as normal samples ##have been collected from GEO before running this function. ##This function may take a long time (> half an hour) to finish. ##Please use R-2.15 and corresponding frma package to run the function. ##Different versions of dependent microarray data preprocessing packages and annotation ##packages may give slightly different results. ##Please contact the author if there is any confusion data("AMC") ge.pre <- geneExpPre(celpath=".", AMC_sample_head) ## End(Not run)## Not run: ##Please make sure that all colon cancer, adenomas as well as normal samples ##have been collected from GEO before running this function. ##This function may take a long time (> half an hour) to finish. ##Please use R-2.15 and corresponding frma package to run the function. ##Different versions of dependent microarray data preprocessing packages and annotation ##packages may give slightly different results. ##Please contact the author if there is any confusion data("AMC") ge.pre <- geneExpPre(celpath=".", AMC_sample_head) ## End(Not run)
This function retrieves centroids of a PAM classifier
getCentroids(fit, data, threshold)getCentroids(fit, data, threshold)
fit |
the fitting result of a PAM classifier to training data. |
data |
the training data used for building the PAM classifier |
threshold |
the threshold for PAM classification |
This is an internal function called by buildClassifier
to retrieve the centroids of PAM classifier built using the AMC-AJCCII-90
data set.
This function will return a numeric matrix of centroids of the PAM classifier.
Xin Wang [email protected]
De Sousa E Melo, F. and Wang, X. and Jansen, M. et al. Poor prognosis colon cancer is defined by a molecularly distinct subtype and precursor lesion. accepted
Tibshirani, Robert and Hastie, Trevor and Narasimhan, Balasubramanian and Chu, Gilbert (2002). Diagnosis of multiple cancer types by shrunken centroids of gene expression. PNAS, 99(10), 6567-6572.
These functions classify and visualize the 90 AMC colon cancer samples.
pamClassify(datsel, signature, pam.rslt, thresh, postRth=1) figClassify(AMC_CRC_clinical, pred, clu.pred, sdat.sig, gclu.f, nam.ord)pamClassify(datsel, signature, pam.rslt, thresh, postRth=1) figClassify(AMC_CRC_clinical, pred, clu.pred, sdat.sig, gclu.f, nam.ord)
datsel |
a numeric value specifying the cutoff of posterior odds in favor of optimal subtype to the other two. |
signature |
a character vector of signature genes. |
pam.rslt |
an list of training results returned by |
thresh |
the selected shrinkage threshold. |
postRth |
a numeric value specifying the cutoff of posterior odds in favor of optimal subtype to the other two. |
AMC_CRC_clinical |
a data frame including include clinical information (details in |
pred |
a numeric matrix of posterior probabilities for samples to be classified to subtypes. |
clu.pred |
a numeric vector of classification labels named by colon cancer samples. |
sdat.sig |
a numeric matrix of expression values of signature genes for the 90 samples. |
gclu.f |
the result of hierarchical clustering on the expression of signature genes for visualization. |
nam.ord |
a character vector of samples ordered by their classification probabilies for visualization. |
The function pamClassify classifies the 90 AMC colon cancer samples
using the gene expression based classifier built by PAM.
The function figClassify generates classification results with a heatmap
of median centred expression values across samples, a track indicating
classification probabilities, as well as a track indicating relapse of patients.
This function pamClassify will return a list including sdat.sig
(a numeric matrix of expression values of signature genes for the 90 samples),
pred (a numeric matrix of posterior probabilities for samples to be
classified to subtypes), nam.ord (a character vector of samples ordered
by their classification probabilies for visualization), gclu.f (result
of hierarchical clustering on the expression of signature genes for
visualization).
Xin Wang [email protected]
De Sousa E Melo, F. and Wang, X. and Jansen, M. et al. Poor prognosis colon cancer is defined by a molecularly distinct subtype and precursor lesion. accepted
data(AMC, package="DeSousa2013") data(dat, package="DeSousa2013") data(uniGenes, package="DeSousa2013") data(diffGenes.f, package="DeSousa2013") data(classifier, package="DeSousa2013") datsel <- sdat[names(uniGenes), ] rownames(datsel) <- uniGenes datsel <- datsel[diffGenes.f, ] pamcl <- pamClassify(datsel, signature, pam.rslt, thresh, postRth=1) figClassify(AMC_CRC_clinical, pamcl$pred, pamcl$clu.pred, pamcl$sdat.sig, pamcl$gclu.f, pamcl$nam.ord)data(AMC, package="DeSousa2013") data(dat, package="DeSousa2013") data(uniGenes, package="DeSousa2013") data(diffGenes.f, package="DeSousa2013") data(classifier, package="DeSousa2013") datsel <- sdat[names(uniGenes), ] rownames(datsel) <- uniGenes datsel <- datsel[diffGenes.f, ] pamcl <- pamClassify(datsel, signature, pam.rslt, thresh, postRth=1) figClassify(AMC_CRC_clinical, pamcl$pred, pamcl$clu.pred, pamcl$sdat.sig, pamcl$gclu.f, pamcl$nam.ord)
This function takes the expression data of top variable (MAD) genes to generate a mapping file between probesets and unique gene symbols.
pbs2unigenes(ge.CRC, sdat)pbs2unigenes(ge.CRC, sdat)
ge.CRC |
a numeric matrix of expression data of genes expressed in at least one sample. |
sdat |
a matrix of median centered expression values of top variable probesets. |
To facilitate the use of the classifier on other platforms, we collapsed the expression levels for probesets to genes. For each gene the probeset with highest overall expression was selected.
This function will save uniGenes (a character vector of unique gene
symbols named by probesets) to a file called uniGenes.RData in the
current working directory.
Xin Wang [email protected]
De Sousa E Melo, F. and Wang, X. and Jansen, M. et al. Poor prognosis colon cancer is defined by a molecularly distinct subtype and precursor lesion. accepted
library(hgu133plus2.db) data(ge.CRC, package="DeSousa2013") data(dat, package="DeSousa2013") ge.CRC <- ge.all[selPbs, ] uniGenes <- pbs2unigenes(ge.CRC, sdat)library(hgu133plus2.db) data(ge.CRC, package="DeSousa2013") data(dat, package="DeSousa2013") ge.CRC <- ge.all[selPbs, ] uniGenes <- pbs2unigenes(ge.CRC, sdat)
These functions performs survival analysis for the AMC data set, and generate a kaplan-meier plot to diagnose the prognosis of three subtypes.
progAMC(AMC_CRC_clinical, AMC_sample_head, clu.pred) figKM(surv, survstats)progAMC(AMC_CRC_clinical, AMC_sample_head, clu.pred) figKM(surv, survstats)
AMC_CRC_clinical |
a data frame including include clinical information (details in |
AMC_sample_head |
a data frame include mapping information between microarray ids and sample ids
(details in |
clu.pred |
a numeric vector of classification labels named by colon cancer samples. |
surv |
the result of |
survstats |
the result of |
The function progAMC performs survival analysis for the AMC data set and
compares the prognosis of the three subtypes.
The function figKM helps visualize the prognosis in a kaplan-meier plot.
This function progAMC will save surv (result of survfit
containing survival curves of the AMC data set), survstats (result of
survdiff), data4surv (the data to perform
survival analysis) to a file called survival.RData in the current
working directory.
Xin Wang [email protected]
De Sousa E Melo, F. and Wang, X. and Jansen, M. et al. Poor prognosis colon cancer is defined by a molecularly distinct subtype and precursor lesion. accepted
library(survival) data(AMC, source="DeSousa2013") data(predAMC, source="DeSousa2013") prog <- progAMC(AMC_CRC_clinical, AMC_sample_head, clu.pred) surv <- prog$surv survstats <- prog$survstats figKM(surv, survstats)library(survival) data(AMC, source="DeSousa2013") data(predAMC, source="DeSousa2013") prog <- progAMC(AMC_CRC_clinical, AMC_sample_head, clu.pred) surv <- prog$surv survstats <- prog$survstats figKM(surv, survstats)
This function selects top variable probesets across the 90 cancer samples by median absolute deviations.
selTopVarGenes(ge.CRC, MADth=0.5)selTopVarGenes(ge.CRC, MADth=0.5)
ge.CRC |
a numeric matrix of expression data of genes expressed in at least one sample. |
MADth |
an numeric value specifying the cutoff of MAD (median absolute deviation). |
This function will return sdat, which is a matrix of median centered
expression values of top variable probesets.
Xin Wang [email protected]
De Sousa E Melo, F. and Wang, X. and Jansen, M. et al. Poor prognosis colon cancer is defined by a molecularly distinct subtype and precursor lesion. accepted
data(ge.CRC, package="DeSousa2013") ge.CRC <- ge.all[selPbs, ] sdat <- selTopVarGenes(ge.CRC, MADth=0.5)data(ge.CRC, package="DeSousa2013") ge.CRC <- ge.all[selPbs, ] sdat <- selTopVarGenes(ge.CRC, MADth=0.5)