| Title: | Detecting Correlated Genomic Regions |
|---|---|
| Description: | Performs correlation matrix segmentation and applies a test procedure to detect highly correlated regions in gene expression. |
| Authors: | Eleni Ioanna Delatola, Emilie Lebarbier, Tristan Mary-Huard, Francois Radvanyi, Stephane Robin, Jennifer Wong |
| Maintainer: | Eleni Ioanna Delatola <[email protected]> |
| License: | GPL-2 |
| Version: | 1.2 |
| Built: | 2024-12-19 06:27:23 UTC |
| Source: | CRAN |
Performs correlation matrix segmentation and applies a test procedure to detect highly correlated regions in gene expression. The segmentation procedure detects changes in the patterns of the gene expression correlation matrix. The test procedure asseses which regions exhibit a significantly high level of correlation. Additionally, a preprocessing procedure is provided to correct gene expression for copy number variation.
| Package: | SegCorr |
| Type: | Package |
| Version: | 1.2 |
| Date: | 2015-01-19 |
| License: | GPL-2 |
E. I. Delatola, E. Lebarbier, T. Mary-Huard, F. Radvanyi, S. Robin, J. Wong.
Maintainer: Eleni Ioanna Delatola <[email protected]>
Delatola E. I., Lebarbier E., Mary-Huard T., Radvanyi F., Robin S., Wong J.(2017). SegCorr: a statistical procedure for the detection of genomic regions of correlated expression. BMC Bioinformatics, 18:333.
#data.sets = c('SNP','EXP_raw') ## Each gene corresponds to one SNP probe ## #Position_EXP = matrix(1:1000,nrow=500,byrow=TRUE) #Position_SNP = seq(2,1000,by=2) #data(list=data.sets) #CHR = rep(1,dim(EXP_raw)[1]) #SNP.CHR = rep(1,dim(SNP)[1]) #results = SegCorr(CHR = CHR, EXP = EXP_raw, CNV = TRUE, SNPSMOOTH=TRUE, #Position.EXP = Position_EXP, SNP.CHR = SNP.CHR, SNP=SNP , Position.SNP = Position_SNP) ################drawing the heatmap for one region ########################### #tau = results$Region.List[1,2]: results$Region.List[1,3] #EXP.CNV = results$EXP.corrected #heatmap(EXP.CNV[tau,])#data.sets = c('SNP','EXP_raw') ## Each gene corresponds to one SNP probe ## #Position_EXP = matrix(1:1000,nrow=500,byrow=TRUE) #Position_SNP = seq(2,1000,by=2) #data(list=data.sets) #CHR = rep(1,dim(EXP_raw)[1]) #SNP.CHR = rep(1,dim(SNP)[1]) #results = SegCorr(CHR = CHR, EXP = EXP_raw, CNV = TRUE, SNPSMOOTH=TRUE, #Position.EXP = Position_EXP, SNP.CHR = SNP.CHR, SNP=SNP , Position.SNP = Position_SNP) ################drawing the heatmap for one region ########################### #tau = results$Region.List[1,2]: results$Region.List[1,3] #EXP.CNV = results$EXP.corrected #heatmap(EXP.CNV[tau,])
Correcting gene expression signal for CNV.
CNV_correction(s.Position.EXP, e.Position.EXP, Position.SNP, mu.SNP, EXP)CNV_correction(s.Position.EXP, e.Position.EXP, Position.SNP, mu.SNP, EXP)
s.Position.EXP |
vector with gene start position |
e.Position.EXP |
vector with gene end position |
Position.SNP |
vector with SNP/CGH positions |
mu.SNP |
Smoothed genomic signal matrix not containing NA values. Rows correspond to probes, while columns to patients. The ordering of the patients must be the same as in the EXP matrix. |
EXP |
Gene expression matrix must not contain NA's and genes with same expression value (i.e. null gene). Rows correspond to probes, while columns to patients. Again, ordering of patients must be the same between EXP and mu.SNP matrices. |
Overlapping genes may correspond to the same SNP/CGH probes.
CNV corrected signal matrix.
E. I. Delatola, E. Lebarbier, T. Mary-Huard, F. Radvanyi, S. Robin, J. Wong.
Delatola E. I., Lebarbier E., Mary-Huard T., Radvanyi F., Robin S., Wong J.(2017). SegCorr: a statistical procedure for the detection of genomic regions of correlated expression. BMC Bioinformatics, 18:333.
#data.sets = c('SNP','EXP_raw') ## Each gene corresponds to one SNP probe ## #Position_EXP = matrix(1:1000,nrow=500,byrow=TRUE) #Position_SNP = seq(2,1000,by=2) #data(list=data.sets) #mu.SNP = segmented_signal(SNP ,100) ## smoothed SNP signal #EXP.CNV = CNV_correction(Position_EXP[,1], Position_EXP[,2], Position_SNP, #mu.SNP, EXP_raw)## corrected signal#data.sets = c('SNP','EXP_raw') ## Each gene corresponds to one SNP probe ## #Position_EXP = matrix(1:1000,nrow=500,byrow=TRUE) #Position_SNP = seq(2,1000,by=2) #data(list=data.sets) #mu.SNP = segmented_signal(SNP ,100) ## smoothed SNP signal #EXP.CNV = CNV_correction(Position_EXP[,1], Position_EXP[,2], Position_SNP, #mu.SNP, EXP_raw)## corrected signal
Gene expression profiles have been generated for 30 patients and 500 genes. Background correlation is set to 0.08 and the correlation for H1 regions to 0.5. The location of the H1 regions is as suggested in the work of Lai et al.(2005), i.e region 1 [101, 105], region 2 [201, 210], region 3 [301, 320] and region 4 [401, 440].
data("EXP_raw")data("EXP_raw")
Data frame containing the gene expression signal for 500 genes (rows) on 30 patients (columns).
Lai, W. R., Johnson, M. D., Kucherlapati, R., & Park, P. J. (2005). Comparative analysis of algorithms for identifying amplifications and deletions in array CGH data. Bioinformatics, 21(19), 3763-3770.
#data(EXP_raw) #G = cor(t(EXP_raw))## calculating the gene x gene correlation matrix #image(G)## plotting the correlation matrix#data(EXP_raw) #G = cor(t(EXP_raw))## calculating the gene x gene correlation matrix #image(G)## plotting the correlation matrix
Gene expression is corrected for CNV events must not contain NA's and genes with same expression value (i.e. null gene expression). Segmentation is used to detect changes in the correlation pattern. Regions with high correlation are identified using an exact test.
SegCorr(CHR, EXP, genes,S, CNV, SNPSMOOTH, Position.EXP, SNP.CHR, SNP, Position.SNP, Kmax)SegCorr(CHR, EXP, genes,S, CNV, SNPSMOOTH, Position.EXP, SNP.CHR, SNP, Position.SNP, Kmax)
CHR |
Chromosome allocation vector for the genes. |
EXP |
Gene expression matrix (raw/corrected for CNV). Columns correspond to patients and rows to genes. The expression matrix must not contain either NA's or genes with same expression value (i.e. null gene expression) |
genes |
Gene ID(name) vector. |
S |
Threshold for model selection. Default S=0.7. |
CNV |
Logical variable indicating whether to perform CNV correction. When CNV=T, the correction is performed. Default value CNV=F. |
SNPSMOOTH |
(Optional Argument when CNV=T) Logical variable indicating whether to perform SNPSMOOTH. When SNPSMOOTH=T, the smoothing is performed. Default value SNPSMOOTH=F. |
Position.EXP |
(Optional Argument when CNV=T) Expression position matrix. First column is the start position and the second is the end position. |
SNP.CHR |
(Optional Argument when CNV=T) Chromosome allocation vector for genomic probes. |
SNP |
(Optional Argument when CNV=T) SNP profile matrix not containing NA's. Columns correspond to patients and rows to probes. |
Position.SNP |
(Optional Argument when CNV=T) vector with SNP positions |
Kmax |
(Optional Argument when CNV=T and SNPSMOOTH=T) Maximum number of segments. (mean profile segmentation) |
Overlapping genes may correspond to the same genomic probes.
Results |
Matrix containing information about the genomic regions. Each region corresponds to a row of the matrix, the one with the smallest p-value is on the top of the list. |
Results$CHR |
Chromosome |
Results$Start/End |
the region boundaries with repsect to the physical location of the gene in the chromosome |
Results$Rho |
|
Results$length |
number of genes in the region |
Results$first/last gene |
name of the first/last gene in the region |
Results$p-value |
p-value as obtained from the test |
Results$genes |
names of the genes belonging to the region |
Results$p-valueadj |
p-value of the region corrected for multiple testing |
Chromosome.Inf |
Matrix containing the estimated background correlation (rho0.hat) per chromsome, the number of segments and the log-loglikehood. |
EXP.corrected |
If the CNV option is chosen, the corrected signal is given. |
E. I. Delatola, E. Lebarbier, T. Mary-Huard, F. Radvanyi, S. Robin, J. Wong.
Delatola E. I., Lebarbier E., Mary-Huard T., Radvanyi F., Robin S., Wong J.(2017). SegCorr: a statistical procedure for the detection of genomic regions of correlated expression. BMC Bioinformatics, 18:333.
#data('EXP_raw') #CHR = rep(1,dim(EXP_raw)[1]) #results = SegCorr(CHR = CHR, EXP = EXP_raw, CNV = FALSE,S=0.7) ################drawing the heatmap for one region ########################### #tau = results$Region.List[1,2]: results$Region.List[1,3] #heatmap(as.matrix(EXP_raw[tau,]))#data('EXP_raw') #CHR = rep(1,dim(EXP_raw)[1]) #results = SegCorr(CHR = CHR, EXP = EXP_raw, CNV = FALSE,S=0.7) ################drawing the heatmap for one region ########################### #tau = results$Region.List[1,2]: results$Region.List[1,3] #heatmap(as.matrix(EXP_raw[tau,]))
For a given chromosome, gene correlation matrix segmentation is performed. Regions with high correlation are identified using an exact test. The expression matrix must not contain NA's and genes with same expression value (i.e. null gene expression).
segmentation(CHR, EXP, genes, S)segmentation(CHR, EXP, genes, S)
CHR |
chromosome name |
EXP |
Gene expression matrix (raw/corrected for CNV). Columns correspond to patients and rows to genes. The expression matrix must not contain either NA's or genes with same expression value (i.e. null gene expression). |
genes |
Gene ID(name) vector. |
S |
Threshold for model selection. Default S=0.7. |
Results |
Matrix containing information about the genomic regions. Each region corresponds to a row of the matrix, the one with the smallest p-value is on the top of the list. |
Results$CHR |
Chromosome |
Results$Start/End |
region boundaries with respect to the physical location of the gene in the chromosome |
Results$Rho |
|
Results$length |
number of genes in the region |
Results$first/last gene |
name of the first/last gene in the region |
Results$p-value |
p-value as obtained from the test |
Results$genes |
names of genes belonging to the region |
rho0 |
estimate of the background correlation |
likelihood |
log-likelihood |
K |
number of segments |
E. I. Delatola, E. Lebarbier, T. Mary-Huard, F. Radvanyi, S. Robin, J. Wong.
Delatola E. I., Lebarbier E., Mary-Huard T., Radvanyi F., Robin S., Wong J.(2017). SegCorr: a statistical procedure for the detection of genomic regions of correlated expression. BMC Bioinformatics, 18:333.
#data(EXP_raw) #G = cor(t(EXP_raw))## calculating the gene x gene correlation matrix #image(G)## plotting the correlation matrix #results = segmentation(EXP = EXP_raw)#data(EXP_raw) #G = cor(t(EXP_raw))## calculating the gene x gene correlation matrix #image(G)## plotting the correlation matrix #results = segmentation(EXP = EXP_raw)
Mean segmentation on the genomic signal is performed using the Fpsn function of the jointseg package.
segmented_signal(SNP.Chr, Kmax)segmented_signal(SNP.Chr, Kmax)
SNP.Chr |
SNP/CGH profile matrix for a given chromosome (NA's not allowed). Columns correspond to patients and rows to probes. |
Kmax |
Maximum number of segments. |
Smoothed genomic signal matrix. Rows correspond to probes and columns to patients.
E. I. Delatola, E. Lebarbier, T. Mary-Huard, F. Radvanyi, S. Robin, J. Wong.
Morgane Pierre-Jean, Guillem Rigaill and Pierre Neuvial. Performance evaluation of DNA copy number segmentation methods. Briefings in Bioinformatics (2015) 16 (4): 600-615.
#data(SNP) #mu.SNP = segmented_signal(SNP ,100)#data(SNP) #mu.SNP = segmented_signal(SNP ,100)
SNP profiles for 30 patients and 500 probes have been simulated. Each gene corresponds to one SNP value.
data("SNP")data("SNP")
Data frame with 500 probes (rows) on 30 patients (patients).
#data(SNP)#data(SNP)