| Title: | LC-MS/MS Differential Expression Tests |
|---|---|
| Description: | Statistical tests for label-free LC-MS/MS data by spectral counts, to discover differentially expressed proteins between two biological conditions. Three tests are available: Poisson GLM regression, quasi-likelihood GLM regression, and the negative binomial of the edgeR package.The three models admit blocking factors to control for nuissance variables.To assure a good level of reproducibility a post-test filter is available, where we may set the minimum effect size considered biologicaly relevant, and the minimum expression of the most abundant condition. |
| Authors: | Josep Gregori, Alex Sanchez, and Josep Villanueva |
| Maintainer: | Josep Gregori i Font <[email protected]> |
| License: | GPL-2 |
| Version: | 1.43.0 |
| Built: | 2024-06-30 06:14:17 UTC |
| Source: | https://github.com/bioc/msmsTests |
Statistical tests for label-free LC-MS/MS data by spectral counts, to discover differentially expressed proteins between two biological conditions. Three tests are available: Poisson GLM regression, quasi-likelihood GLM regression, and the negative binomial of the edgeR package. The three models admit blocking factors to control for nuissance variables. To assure a good level of reproducibility a post-test filter is available, where we may set the minimum effect size considered biologicaly relevant, and the minimum expression of the most abundant condition.
| Package: | msmsTests |
| Type: | Package |
| Version: | 0.99.1 |
| Date: | 2013-07-26 |
| License: | GPL-2 |
msms.glm.pois: |
Poisson based GLM regression |
msms.glm.qlll: |
Quasi-likelihood GLMregression |
msms.edgeR: |
The binomial negative of edgeR |
pval.by.fc: |
Table of cumulative frequencies of features by p-values in bins of log fold change |
test.results: |
Multitest p-value adjustement and post-test filter |
res.volcanoplot: |
Volcanplot of the results |
Josep Gregori, Alex Sanchez, and Josep Villanueva
Maintainer: Josep Gregori <[email protected]>
Josep Gregori, Laura Villareal, Alex Sanchez, Jose Baselga, Josep Villanueva (2013). An Effect Size Filter Improves the Reproducibility in Spectral Counting-based Comparative Proteomics. Journal of Proteomics, DOI http://dx.doi.org/10.1016/j.jprot.2013.05.030
Given a null and an alternative model, with a two level treatment factor as the two conditions to compare, executes the negative binomial test by edgeR functions to discover differentially expressed proteins between the two conditions. The null and alternative models may include blocking factors.The reference level of the main factor is considered to be the control condition
msms.edgeR(msnset,form1,form0,facs=NULL,div=NULL,fnm=NULL)msms.edgeR(msnset,form1,form0,facs=NULL,div=NULL,fnm=NULL)
msnset |
A MSnSet object with spectral counts in the expression matrix. |
form1 |
The alternative hypothesis model as an standard R formula, with the treatment factor of interest, and eventual blocking factors. |
form0 |
The null hypothesis model as an standard R formula.It may be the standard null model (y~.) or contain one or multiple blocking factors. |
facs |
NULL or a data frame with the factors in its columns. |
div |
NULL or a vector with the divisors used to compute the offsets. |
fnm |
NULL or a character string with the treatment factor name, as used in the column names of the factors data frame, and in the formula. |
The right hand site of the formulas is expected to be "y~", with
the combination of factors after the tilde. If facs is NULL the factors
are taken as default from pData(msnset). If div is NULL all
divisors are taken equal to one. If fnm is NULL it is taken to be the
first factor in facs.
A data frame with column names 'LogFC', 'LR', 'p.value', with the estimated log fold changes, likelihood ratio statistic and corresponding p-value as obtaimed from a call to glmLRT() from the edgeR package.
Josep Gregori i Font
Robinson MD, McCarthy DJ and Smyth GK (2010). edgeR: a Bioconductor package for differential expression analysis of digital gene expression data. Bioinformatics 26, 139-140
Robinson MD and Smyth GK (2007). Moderated statistical tests for assessing differences in tag abundance. Bioinformatics 23, 2881-2887
Robinson MD and Smyth GK (2008). Small-sample estimation of negative binomial dispersion, with applications to SAGE data. Biostatistics, 9, 321-332
Josep Gregori, Laura Villareal, Alex Sanchez, Jose Baselga, Josep Villanueva (2013). An Effect Size Filter Improves the Reproducibility in Spectral Counting-based Comparative Proteomics. Journal of Proteomics, DOI http://dx.doi.org/10.1016/j.jprot.2013.05.030
MSnSet, edgeR, glmLRT, msmsEDA
## Example library(msmsTests) data(msms.dataset) e <- pp.msms.data(msms.dataset) e null.f <- "y~batch" alt.f <- "y~treat+batch" div <- apply(exprs(e),2,sum) res <- msms.edgeR(e,alt.f,null.f,div=div,fnm="treat") str(res) head(res)## Example library(msmsTests) data(msms.dataset) e <- pp.msms.data(msms.dataset) e null.f <- "y~batch" alt.f <- "y~treat+batch" div <- apply(exprs(e),2,sum) res <- msms.edgeR(e,alt.f,null.f,div=div,fnm="treat") str(res) head(res)
Given a null and an alternative model, with a two level treatment factor as the two conditions to compare, executes a Poisson based GLM regression to discover differentially expressed proteins between the two conditions. The null and alternative models may include blocking factors.The reference level of the main factor is considered to be the control condition.
msms.glm.pois(msnset,form1,form0,facs=NULL,div=NULL)msms.glm.pois(msnset,form1,form0,facs=NULL,div=NULL)
msnset |
A MSnSet object with spectral counts in the expression matrix. |
form1 |
The alternative hypothesis model as an standard R formula, with the treatment factor of interest, and eventual blocking factors. |
form0 |
The null hypothesis model as an standard R formula.It may be the standard null model (y~.) or contain one or multiple blocking factors. |
facs |
NULL or a data frame with the factors in its columns. |
div |
NULL or a vector with the divisors used to compute the offsets. |
The right hand site of the formulas is expected to be "y~", with
the combination of factors after the tilde. If facs is NULL the factors
are taken as default from pData(msnset). If div is NULL all
divisors are taken equal to one.
A data frame with the following columns:
LogFC |
Log fold change estimated from the model parameters. |
D |
Residual deviance as statistic of the test. |
p.value |
The p-values obtained from the test. |
Josep Gregori i Font
Agresti, A. (2002) Categorical Data Analysis, 2nd Edition, John Wiley & Sons, Inc., Hoboken, New Jersey
Thompson L.A. (2009) R (and S-PLUS) Manual to Accompany Agresti s Categorical Data Analysis (2002), 2nd edition https://home.comcast.net/~lthompson221/Splusdiscrete2.pdf
Dobson, A.J. (2002) An Introduction to Generalized Linear Models, 2nd Edition, Chapman & Hall/CRC, New York
Venables, W. N. and Ripley, B. D. (2002) Modern Applied Statistics with S. New York: Springer
Josep Gregori, Laura Villareal, Alex Sanchez, Jose Baselga, Josep Villanueva (2013). An Effect Size Filter Improves the Reproducibility in Spectral Counting-based Comparative Proteomics. Journal of Proteomics, DOI http://dx.doi.org/10.1016/j.jprot.2013.05.030
library(msmsTests) data(msms.dataset) # Pre-process expression matrix e <- pp.msms.data(msms.dataset) # Factors pData(e) # Control condition levels(pData(e)$treat)[1] # Treatment condition levels(pData(e)$treat)[2] # Models and normalizing condition null.f <- "y~batch" alt.f <- "y~treat+batch" div <- apply(exprs(e),2,sum) #Test res <- msms.glm.pois(e,alt.f,null.f,div=div) str(res) head(res)library(msmsTests) data(msms.dataset) # Pre-process expression matrix e <- pp.msms.data(msms.dataset) # Factors pData(e) # Control condition levels(pData(e)$treat)[1] # Treatment condition levels(pData(e)$treat)[2] # Models and normalizing condition null.f <- "y~batch" alt.f <- "y~treat+batch" div <- apply(exprs(e),2,sum) #Test res <- msms.glm.pois(e,alt.f,null.f,div=div) str(res) head(res)
Given a null and an alternative model, with a two level treatment factor as the two conditions to compare, executes a quasi-likelihood based GLM regression to discover differentially expressed proteins between the two conditions. The null and alternative models may include blocking factors.The reference level of the main factor is considered to be the control condition.
msms.glm.qlll(msnset,form1,form0,facs=NULL,div=NULL)msms.glm.qlll(msnset,form1,form0,facs=NULL,div=NULL)
msnset |
A MSnSet object with spectral counts in the expression matrix. |
form1 |
The alternative hypothesis model as an standard R formula, with the treatment factor of interest, and eventual blocking factors. |
form0 |
The null hypothesis model as an standard R formula.It may be the standard null model (y~.) or contain one or multiple blocking factors. |
facs |
NULL or a data frame with the factors in its columns. |
div |
NULL or a vector with the divisors used to compute the offsets. |
The right hand site of the formulas is expected to be "y~", with
the combination of factors after the tilde. If facs is NULL the factors
are taken as default from pData(msnset). If div is NULL all
divisors are taken equal to one.
A data frame with the following columns:
LogFC |
Log fold change estimated from the model parameters. |
D |
Residual deviance as statistic of the test. |
p.value |
The p-values obtained from the test. |
Josep Gregori i Font
Agresti, A. (2002) Categorical Data Analysis, 2nd Edition, John Wiley & Sons, Inc., Hoboken, New Jersey
Thompson L.A. (2009) R (and S-PLUS) Manual to Accompany Agresti s Categorical Data Analysis, 2nd edition https://home.comcast.net/~lthompson221/Splusdiscrete2.pdf
Dobson, A.J. (2002) An Introduction to Generalized Linear Models, 2nd Edition, Chapman & Hall/CRC, New York
Venables, W. N. and Ripley, B. D. (2002) Modern Applied Statistics with S, New York: Springer
Li, M.; Gray, W.; Zhang, H.; Chung, C. H.; Billheimer, D.; Yarbrough, W. G.; Liebler, D. C.; Shyr, Y.; Slebos, R. J. C. Comparative shotgun proteomics using spectral count data and quasi-likelihood modeling, J Proteome Res 2010, 9, 4295-4305.
Josep Gregori, Laura Villareal, Alex Sanchez, Jose Baselga, Josep Villanueva (2013). An Effect Size Filter Improves the Reproducibility in Spectral Counting-based Comparative Proteomics, Journal of Proteomics, http://dx.doi.org/10.1016/j.jprot.2013.05.030
library(msmsTests) data(msms.dataset) # Pre-process expression matrix e <- pp.msms.data(msms.dataset) # Factors pData(e) # Control condition levels(pData(e)$treat)[1] # Treatment condition levels(pData(e)$treat)[2] # Models and normalizing condition null.f <- "y~batch" alt.f <- "y~treat+batch" div <- apply(exprs(e),2,sum) #Test res <- msms.glm.qlll(e,alt.f,null.f,div=div) str(res) head(res)library(msmsTests) data(msms.dataset) # Pre-process expression matrix e <- pp.msms.data(msms.dataset) # Factors pData(e) # Control condition levels(pData(e)$treat)[1] # Treatment condition levels(pData(e)$treat)[2] # Models and normalizing condition null.f <- "y~batch" alt.f <- "y~treat+batch" div <- apply(exprs(e),2,sum) #Test res <- msms.glm.qlll(e,alt.f,null.f,div=div) str(res) head(res)
A MSnSet with a spectral counts in the expression matrix and a treatment factor
in the phenoData slot.
The spectral counts matrix has samples in the columns, and proteins in the rows.
Each sample consists in 500ng of standard yeast lisate spiked with 100, 200, 400 and 600fm of a mix of 48 equimolar human proteins (UPS1, Sigma-Aldrich). The dataset contains a different number of technical replicates of each sample.
data(msms.spk)data(msms.spk)
A MSnSet
Josep Gregori, Laura Villareal, Alex Sanchez, Jose Baselga, Josep Villanueva (2013). An Effect Size Filter Improves the Reproducibility in Spectral Counting-based Comparative Proteomics. Journal of Proteomics, DOI http://dx.doi.org/10.1016/j.jprot.2013.05.030
Laurent Gatto and Kathryn S. Lilley, MSnbase - an R/Bioconductor package for isobaric tagged mass spectrometry data visualization, processing and quantitation, Bioinformatics 28(2), 288-289 (2012).
See MSnSet for detail on the class, and the exprs and pData accessors.
data(msms.spk) msms.spk dim(msms.spk) table(pData(msms.spk)) head(exprs(msms.spk))data(msms.spk) msms.spk dim(msms.spk) table(pData(msms.spk)) head(exprs(msms.spk))
Given the set of p-values and log fold changes that result from a test, computes a table of cumulative frequencies of features by p-values in bins of log fold changes.
pval.by.fc(pvals,lfc)pval.by.fc(pvals,lfc)
lfc |
The log fold changes estimated from the tests. |
pvals |
The p-values, adjusted or not, obtained from the tests. |
A matrix of cumulated frequencies with descriptive row and column names.
Josep Gregori i Font
Josep Gregori, Laura Villareal, Alex Sanchez, Jose Baselga, Josep Villanueva (2013). An Effect Size Filter Improves the Reproducibility in Spectral Counting-based Comparative Proteomics. Journal of Proteomics, DOI http://dx.doi.org/10.1016/j.jprot.2013.05.030
library(msmsTests) data(msms.spk) # Subset treat <- pData(msms.spk) jdx <- which(treat=="U200" | treat=="U600") e <- msms.spk[,jdx] pData(e)$treat <- treat[jdx,1,drop=TRUE] # Pre-process expression matrix e <- pp.msms.data(e) # Models and normalizing condition null.f <- "y~1" alt.f <- "y~treat" div <- apply(exprs(e),2,sum) #Test res <- msms.glm.pois(e,alt.f,null.f,div=div) # Post-test filter lst <- test.results(res,e,pData(e)$treat,"U600","U200",div, alpha=0.05,minSpC=2,minLFC=1, method="BH") ## On all features, with multitest adjusted p-values pval.by.fc(lst$tres$adjp, lst$tres$LogFC) ### On all features deemed significant and biologically relevant flags <- lst$tres$DEP pval.by.fc(lst$tres$adjp[flags], lst$tres$LogFC[flags])library(msmsTests) data(msms.spk) # Subset treat <- pData(msms.spk) jdx <- which(treat=="U200" | treat=="U600") e <- msms.spk[,jdx] pData(e)$treat <- treat[jdx,1,drop=TRUE] # Pre-process expression matrix e <- pp.msms.data(e) # Models and normalizing condition null.f <- "y~1" alt.f <- "y~treat" div <- apply(exprs(e),2,sum) #Test res <- msms.glm.pois(e,alt.f,null.f,div=div) # Post-test filter lst <- test.results(res,e,pData(e)$treat,"U600","U200",div, alpha=0.05,minSpC=2,minLFC=1, method="BH") ## On all features, with multitest adjusted p-values pval.by.fc(lst$tres$adjp, lst$tres$LogFC) ### On all features deemed significant and biologically relevant flags <- lst$tres$DEP pval.by.fc(lst$tres$adjp[flags], lst$tres$LogFC[flags])
Given the data frame obtained from test.results() a volcano
plot is drawn.The features are colored according to significance and
relevance.
res.volcanoplot(tres, max.pval=0.05, min.LFC=1, maxx=3, maxy=10, ylbls=20)res.volcanoplot(tres, max.pval=0.05, min.LFC=1, maxx=3, maxy=10, ylbls=20)
tres |
The dataframe with test results as obtained from |
max.pval |
The maximum adjusted p-value considered as statistically significant. |
min.LFC |
The minimum absolute log fold change considered as biologically relevant. |
maxx |
The maximum value in abcissas (i.e. log2(fold change)). |
maxy |
The maximum value in ordinates (i.e. -log10(p.val)) |
ylbls |
All features with -log10(p.val) above this value will be ploted with feature labels. |
Abscissas and ordinates may be limited giving a value other than NULL to the
parameters maxx and maxy. All features deemed significant and
relevant are ploted by a blue dot, all features deemed significant but not
passing the post test filter are plotted by a red dot. The non-significant
features are plotted as smaller black dots. All features deemed significant
and relevant and with a -log10 p-value above ylbls are plotted with a
label showing their row index in the test results dataframe.
The borders limiting the values given by max.pval and min.LFC
are ploted as dash-and-dot red lines.
No return value.
Josep Gregori i Font
Josep Gregori, Laura Villareal, Alex Sanchez, Jose Baselga, Josep Villanueva (2013). An Effect Size Filter Improves the Reproducibility in Spectral Counting-based Comparative Proteomics. Journal of Proteomics, DOI http://dx.doi.org/10.1016/j.jprot.2013.05.030
library(msmsTests) data(msms.dataset) # Pre-process expression matrix e <- pp.msms.data(msms.dataset) # Models and normalizing condition null.f <- "y~batch" alt.f <- "y~treat+batch" div <- apply(exprs(e),2,sum) #Test res <- msms.glm.qlll(e,alt.f,null.f,div=div) lst <- test.results(res,e,pData(e)$treat,"U600","U200",div, alpha=0.05,minSpC=2,minLFC=log2(1.8), method="BH") # Plot res.volcanoplot(lst$tres, max.pval=0.05, min.LFC=1, maxx=3, maxy=NULL, ylbls=4)library(msmsTests) data(msms.dataset) # Pre-process expression matrix e <- pp.msms.data(msms.dataset) # Models and normalizing condition null.f <- "y~batch" alt.f <- "y~treat+batch" div <- apply(exprs(e),2,sum) #Test res <- msms.glm.qlll(e,alt.f,null.f,div=div) lst <- test.results(res,e,pData(e)$treat,"U600","U200",div, alpha=0.05,minSpC=2,minLFC=log2(1.8), method="BH") # Plot res.volcanoplot(lst$tres, max.pval=0.05, min.LFC=1, maxx=3, maxy=NULL, ylbls=4)
Operates on the statistic tests results obtained from msms.glm.pois(),
msms.glm.qlll() or msms.edgeR(). The following variables are
computed: Raw expression mean values for each condition (control and treatment),
log fold change based on these expression levels and taking into account the
normalizing divisors (div), multitest adjusted p-values with FDR control,
and a post test filter based on minimum spectral counts and minimum absolute log
fold change as estimated by the statistic test. According to the results of this
post-test filter, features are flagged as T or F depending on whether they
result relevant or not, beyond their statistic signicance.
test.results(test, msnset, gpf, gp1, gp2, div, alpha = 0.05, minSpC = 2, minLFC = 1, method = "BH")test.results(test, msnset, gpf, gp1, gp2, div, alpha = 0.05, minSpC = 2, minLFC = 1, method = "BH")
test |
The dataframe obtained from either |
msnset |
A MSnSet object with spectral counts in the expression matrix. |
gpf |
The factor used in the tests. |
gp1 |
The treatment level name. |
gp2 |
The control level name. Should be the factor's reference level.
See R function |
div |
The weights used as divisors (offsets) in the GLM model. Usually the sum of spectral counts of each sample. |
alpha |
The multi test adjusted p-value significance threshold. |
minSpC |
The minimum spectral counts considered as relevant in the most abundant condition. This filter aims at reaching good reproducibility. |
minLFC |
The minimum absolute log fold change considered both, relevant and biologically significant. This filter aims at assuring enough biological effect size and at reaching good reproducibility. |
method |
One among |
No feature is removed in the filter, but instead they are flagged as TRUE or FALSE
depending on whether they are considered as differentially expressed or not, in the DEP column, taking into account statistic significance and
reproducibility metrics.
A data frame with the following columns:
first column |
Column named as the treatment level with the mean raw spectral counts observed for this condition |
second column |
Column named as the control level with the mean raw spectral counts observed for this condition |
lFC.Av |
Log fold change computed from the mean expression levesl taking into account the given normalization factors. |
logFC |
Log fold change estimated by fitting the given GLM model. The reference level of the main factor is taken as control. |
D or LR |
The statistic obtained from the tests. The residual deviance
|
p.val |
The unadjusted p-values obtained from the tests. |
adjp |
The multitest adjusted p-values with FDR control. |
DEP |
A logical flagging the features considered both as statistically significant and relevant for reproducibility. |
Josep Gregori i Font
Josep Gregori, Laura Villareal, Alex Sanchez, Jose Baselga, Josep Villanueva (2013). An Effect Size Filter Improves the Reproducibility in Spectral Counting-based Comparative Proteomics. Journal of Proteomics, DOI http://dx.doi.org/10.1016/j.jprot.2013.05.030
Benjamini, Y., and Hochberg, Y. (1995). Controlling the false discovery rate: a practical and powerful approach to multiple testing. Journal of the Royal Statistical Society Series B, 57, 289-300.
Alan Dabney, John D. Storey and with assistance from Gregory R. Warnes. qvalue: Q-value estimation for false discovery rate control. R package version 1.30.0.
library(msmsTests) data(msms.dataset) # Pre-process expression matrix e <- pp.msms.data(msms.dataset) # Factors pData(e) # Control condition levels(pData(e)$treat)[1] # Treatment condition levels(pData(e)$treat)[2] # Models and normalizing condition null.f <- "y~batch" alt.f <- "y~treat+batch" div <- apply(exprs(e),2,sum) #Test res <- msms.glm.qlll(e,alt.f,null.f,div=div) # Post-test filter lst <- test.results(res,e,pData(e)$treat,"U600","U200",div, alpha=0.05,minSpC=2,minLFC=1, method="BH") str(lst) lst$cond head(lst$tres) rownames(lst$tres)[which(lst$tres$DEP)]library(msmsTests) data(msms.dataset) # Pre-process expression matrix e <- pp.msms.data(msms.dataset) # Factors pData(e) # Control condition levels(pData(e)$treat)[1] # Treatment condition levels(pData(e)$treat)[2] # Models and normalizing condition null.f <- "y~batch" alt.f <- "y~treat+batch" div <- apply(exprs(e),2,sum) #Test res <- msms.glm.qlll(e,alt.f,null.f,div=div) # Post-test filter lst <- test.results(res,e,pData(e)$treat,"U600","U200",div, alpha=0.05,minSpC=2,minLFC=1, method="BH") str(lst) lst$cond head(lst$tres) rownames(lst$tres)[which(lst$tres$DEP)]