MSstats (version 3.4.0)

dataProcess: Data pre-processing and quality control of MS runs of raw data

Description

Data pre-processing and quality control of MS runs of the original raw data into quantitative data for model fitting and group comparison. Log transformation is automatically applied and additional variables are created in columns for model fitting and group comparison process. Three options of data pre-processing and quality control of MS runs in dataProcess are (1) Transformation: logarithm transformation with base 2 or 10; (2) Normalization: to remove systematic bias between MS runs.

Usage

dataProcess(raw,logTrans=2,
		normalization="equalizeMedians",nameStandards=NULL,
		betweenRunInterferenceScore=FALSE, address="",
		fillIncompleteRows=TRUE,
		featureSubset="all",
		remove_proteins_with_interference=FALSE, n_top_feature=3,
		summaryMethod="TMP",
		equalFeatureVar=TRUE,
		filterLogOfSum=TRUE,
		censoredInt="NA",
		cutoffCensored="minFeature",
		MBimpute=TRUE,
		original_scale=FALSE,
		logsum=FALSE,
		remove50missing=FALSE,
		skylineReport=FALSE)

Arguments

raw

name of the raw (input) data set.

logTrans

logarithm transformation with base 2(default) or 10.

normalization

normalization to remove systematic bias between MS runs. There are three different normalizations supported. 'equalizeMedians'(default) represents constant normalization (equalizing the medians) based on reference signals is performed. 'quantile' represents quantile normalization based on reference signals is performed. 'globalStandards' represents normalization with global standards proteins. FALSE represents no normalization is performed.

nameStandards

vector of global standard peptide names. only for normalization with global standard peptides.

betweenRunInterferenceScore

interference is detected by a between-run-interference score. TRUE means the scores are generated automatically and stored in a .csv file. FALSE(default) means no scores are generated.

fillIncompleteRows

If the input dataset has incomplete rows, TRUE(default) adds the rows with intensity value=NA for missing peaks. FALSE reports error message with list of features which have incomplete rows.

featureSubset

"all"(default) uses all features that the data set has. "top3" uses top 3 features which have highest average of log2(intensity) across runs. "topN" uses top N features which has highest average of log2(intensity) across runs. It needs the input for n_top_feature option. "highQuality" selects the most informative features which agree the pattern of the average features across the runs.

remove_proteins_with_interference

TRUE allows the algorithm to remove the proteins if deem interfered. FALSE (default) does not allow to remove the proteins, in which all features are interfered. In this case, the proteins, which will completely loss all features by the algorithm, will keep the most abundant peptide.

n_top_feature

The number of top features for featureSubset='topN'. Default is 3, which means to use top 3 features.

summaryMethod

"TMP"(default) means Tukey's median polish, which is robust estimation method. "linear" uses linear mixed model. "logOfSum" conducts log2 (sum of intensities) per run.

equalFeatureVar

only for summaryMethod="linear". default is TRUE. Logical variable for whether the model should account for heterogeneous variation among intensities from different features. Default is TRUE, which assume equal variance among intensities from features. FALSE means that we cannot assume equal variance among intensities from features, then we will account for heterogeneous variation from different features.

filterLogOfSum

For summaryMethod="logOfSum" option, TRUE (default) will filter out the runs which have any missing value. FALSE will not remove any run or features.

censoredInt

Missing values are censored or at random. 'NA' (default) assumes that all 'NA's in 'Intensity' column are censored. '0' uses zero intensities as censored intensity. In this case, NA intensities are missing at random. The output from Skyline should use '0'. Null assumes that all NA intensites are randomly missing.

cutoffCensored

Cutoff value for censoring. only with censoredInt='NA' or '0'. Default is 'minFeature', which uses minimum value for each feature.'minFeatureNRun' uses the smallest between minimum value of corresponding feature and minimum value of corresponding run. 'minRun' uses minumum value for each run.

MBimpute

only for summaryMethod="TMP" and censoredInt='NA' or '0'. TRUE (default) imputes 'NA' or '0' (depending on censoredInt option) by Accelated failure model. FALSE uses the values assigned by cutoffCensored.

original_scale

Default is FALSE, which uses log transformed intensity for TMP. TRUE uses original intensitiesy for TMP.

logsum

Default is FALSE, which uses log of sum intensities after residuals from TMP.

remove50missing

only for summaryMethod="TMP". TRUE removes the runs which have more than 50% missing values. FALSE is default.

skylineReport

default is FALSE. 'TRUE' means raw (input) data set from Skyline MSstats input format, which includes 'Truncated' column and can distinguish zero value and NA (missing values). Zero values in 'Intensity' column will be kept for 'skyline' summary method. Otherwise, they will be replaced with one in order to log transform.

address

the name of folder that will store the results. Default folder is the current working directory. The other assigned folder has to be existed under the current working directory. An output csv file is automatically created with the default name of "BetweenRunInterferenceFile.csv". The command address can help to specify where to store the file as well as how to modify the beginning of the file name.

Warning

When a transition is missing completely in a condition or a MS run, a warning message is sent to the console notifying the user of the missing transitions. The types of experiment that MSstats can analyze are LC-MS, SRM, DIA(SWATH) with label-free or labeled synthetic peptides. MSstats does not support for metabolic labeling or iTRAQ experiments.

Details

• raw : See SRMRawData for the required data structure of raw (input) data.
• logTrans : if logTrans=2, the measurement of Variable ABUNDANCE is log-transformed with base 2. Same apply to logTrans=10.
• normalization : if normalization=TRUE and logTrans=2, the measurement of Variable ABUNDANCE is log-transformed with base 2 and normalized. Same as for logTrans=10.
• featureSubset : After the data was normalized, we deeply looked at each single feature (which is a precursor in DDA, a fragment in DIA, and a transition in SRM) and quantify its un-explainable variation. Ultimately, we remove the features with interference.
• equalFeatureVar : If the unequal variation of error for different peptide features is detected, then a possible solution is to account for the unequal error variation by means of a procedure called iteratively re-weighted least squares. equalFeatureVar=FALSE performs an iterative fitting procedure, in which features are weighted inversely proportionaly to the variation in their intensities, so that feature with large variation are given less importance in the estimation of parameters in the model.

References

Meena Choi, Ching-Yun Chang, Timothy Clough, Daniel Broudy, Trevor Killeen, Brendan MacLean and Olga Vitek. "MSstats: an R package for statistical analysis of quantitative mass spectrometry-based proteomic experiments" Bioinformatics, 30(17):2524-2526, 2014.

Ching-Yun Chang, Paola Picotti, Ruth Huttenhain, Viola Heinzelmann-Schwarz, Marko Jovanovic, Ruedi Aebersold, Olga Vitek. "Protein significance analysis in selected reaction monitoring (SRM) measurements" Molecular & Cellular Proteomics, 11:M111.014662, 2012.

Timothy Clough, Safia Thaminy, Susanne Ragg, Ruedi Aebersold, Olga Vitek. "Statistical protein quantification and significance analysis in label-free LC-M experiments with complex designs" BMC Bioinformatics, 13:S16, 2012.

Examples

# Consider a raw data (i.e. SRMRawData) for a label-based SRM experiment from a yeast study with ten time points (T1-T10) of interests and three biological replicates. 
# It is a time course experiment. The goal is to detect protein abundance changes across time points.

head(SRMRawData)

# Log2 transformation and normalization are applied (default)
QuantData<-dataProcess(SRMRawData)
head(QuantData$ProcessedData)

# Log10 transformation and normalization are applied
QuantData1<-dataProcess(SRMRawData, logTrans=10)
head(QuantData1$ProcessedData)

# Log2 transformation and no normalization are applied
QuantData2<-dataProcess(SRMRawData,normalization=FALSE)
head(QuantData2$ProcessedData)