The MSstatsConvert
package is a member of the MSstatst
family of packages, MSstats and MSstatsTMT. It creates an abstraction for the steps in mass spectrometry (MS) data analysis that are required before a dataset can be used for statistical modeling. In short, the package is responsible for converting output from signal processing tools such as OpenMS
or MaxQuant
into a format suitable for statistical analysis. This includes:
MSstatsConvert
allows for transforming any MS quantification result into a format required by MSstats
and MSstatsTMT
packages. Additionally, it provides built-in cleaning functions for outputs of DIAUmpire
, MaxQuant
, OpenMS
, OpenSWATH
, Progenesis
, ProteomeDiscoverer
, Skyline
, Spectromine
, and Spectronaut
. These functions serve as a base for converter functions (called *toMSstatsFormat
or *toMSstatsTMTFormat
) provided by the MSstats
and MSstatsTMT
packages.
MSstats family packages works with label-free, SRM and TMT datasets. The following column are required.
ProteinName
: column that indicates a protein ID. If the analysis is to be made at the peptide-level, the column should store peptide IDs. Summarization performed by MSstats
and MSstatsTMT
packages is done separately for each ID in this column,PeptideSequence
, PrecursorCharge
, FragmentIon
and ProductCharge
: these four columns define a spectral feature (transition in SRM case). If information for any of the columns is not available, it should be set to a constant value (for example NA
),IsotopeLabelType
: column that indicates whether the measurement is based on an endogenous peptide (indicated by value “L” or “light”) or reference peptide (indicated by value “H” or “heavy”),Run
: column that stores IDs of mass spectrometry runs. If annotation describing biological conditions and replicates is provided via a separate table, the run IDs should match Run
IDs in the annotation,Condition
: column that stores labels for biological conditions (groups of interest). For time-course experiments, this column will represent time points. If the design experiment includes both time points and distinct biological subjects, these labels should be a combination of subject and time point,BioReplicate
: this column should contain a unique identifier for each biological replicate in the experiment. For example, in a clinical proteomic investigation this should be a unique patient ID. Patients from distinct groups (indicated by the Condition
column) should have distinct IDs,Intensity
: column that stores untransformed (in particular, no log transformation) measurements of feature abundance in a given Run (and Channel in TMT case). They can be peak heights, peak areas under the curve, or other quantitative representations of feature abundance,Channel
column is required. Similarly to the Run
column, values in this column must correspond to values in the annotation file, if provided separately.Additionally, if the experiment involves fractionation, Fraction
column can be added to store fraction labels.
MSstatsConvert
allows for flexible logging based on the log4r
package. Information about preprocessing steps can be written to a file, to a console, to both or to neither. The MSstatsLogsSettings
function helps manage log settings. The user can pass a path to an existing file to the log_file_path
parameter. Combined with setting append = TRUE
, this allows writing all information related to a specific data analysis to a single file. If a user does not specify a file, a new file will be created automatically with a name starting with “MSstats_log”, followed by a timestamp.
library(MSstatsConvert)
# default - creates a new file
MSstatsLogsSettings(use_log_file = TRUE, append = FALSE)
# default - creates a new file
MSstatsLogsSettings(use_log_file = TRUE, append = TRUE,
log_file_path = "log_file.log")
# switches logging off
MSstatsLogsSettings(use_log_file = FALSE, append = FALSE)
# switches off logs and messages
MSstatsLogsSettings(use_log_file = FALSE, verbose = FALSE)
Additionally, session info generated by the utils::sessionInfo()
function can be saved to file with the MSstatsSaveSessionInfo
function.
By default, the output file name will start with “MSstats_session_info” and end with a current timestamp.
MS data processing by MSstatsConvert
starts with importing and cleaning data. The MSstatsImport
function produces a wrapper for possibly multiple files that may describe a single dataset. For example, MaxQuant
output consists of two files, while OpenMS
outputs just a single file.
maxquant_evidence = read.csv(system.file("tinytest/raw_data/MaxQuant/mq_ev.csv",
package = "MSstatsConvert"))
maxquant_proteins = read.csv(system.file("tinytest/raw_data/MaxQuant/mq_pg.csv",
package = "MSstatsConvert"))
maxquant_imported = MSstatsImport(list(evidence = maxquant_evidence,
protein_groups = maxquant_proteins),
type = "MSstats", tool = "MaxQuant")
is(maxquant_imported)
#> [1] "MSstatsMaxQuantFiles" "MSstatsInputFiles"
openms_input = read.csv(system.file(
"tinytest/raw_data/OpenMSTMT/openmstmt_input.csv",
package = "MSstatsConvert"
))
openms_imported = MSstatsImport(list(input = openms_input),
"MSstatsTMT", "OpenMS")
is(openms_imported)
#> [1] "MSstatsOpenMSFiles" "MSstatsInputFiles"
The getInputFile
method allows user to retrieve the files:
getInputFile(maxquant_imported, "evidence")[1:5, 1:5]
#> Sequence Length Modifications Modifiedsequence OxidationMProbabilities
#> 1: AEAPAAAPAAK 11 Unmodified _AEAPAAAPAAK_
#> 2: AEAPAAAPAAK 11 Unmodified _AEAPAAAPAAK_
#> 3: AEAPAAAPAAK 11 Unmodified _AEAPAAAPAAK_
#> 4: AEAPAAAPAAK 11 Unmodified _AEAPAAAPAAK_
#> 5: AEAPAAAPAAK 11 Unmodified _AEAPAAAPAAK_
As a next step of the analysis, input files are combined into a single data.table
with standardized column names by the MSstatsClean
function. It is a generic function with built-in support for outputs of tools listed in the “Purpose of the MSstatsConvert package” section. The type
parameter is equal to either MSstats
or MSstatsTMT
and indicates if the data comes from a labelled TMT experiment.
For some datasets, MSstatsClean
may require additional parameters described in the respective help files. For our example datasets, the following calls merge input files into a single table.
maxquant_cleaned = MSstatsClean(maxquant_imported, protein_id_col = "Proteins")
head(maxquant_cleaned)
#> ProteinName PeptideSequence Modifications PrecursorCharge
#> 1: P06959 AEAPAAAPAAK Unmodified 2
#> 2: P06959 AEAPAAAPAAK Unmodified 2
#> 3: P06959 AEAPAAAPAAK Unmodified 2
#> 4: P06959 AEAPAAAPAAK Unmodified 2
#> 5: P06959 AEAPAAAPAAK Unmodified 2
#> 6: P06959 AEAPAAAPAAK Unmodified 2
#> Run Intensity Score
#> 1: 121219_S_CCES_01_01_LysC_Try_1to10_Mixt_1_1 4023100 76.332
#> 2: 121219_S_CCES_01_02_LysC_Try_1to10_Mixt_1_2 5132500 83.081
#> 3: 121219_S_CCES_01_03_LysC_Try_1to10_Mixt_1_3 2761600 104.430
#> 4: 121219_S_CCES_01_05_LysC_Try_1to10_Mixt_2_2 4091800 94.465
#> 5: 121219_S_CCES_01_06_LysC_Try_1to10_Mixt_2_3 4727000 88.596
#> 6: 121219_S_CCES_01_08_LysC_Try_1to10_Mixt_3_2 2258400 90.050
openms_cleaned = MSstatsClean(openms_imported)
head(openms_cleaned)
#> ProteinName PeptideSequence
#> 1: sp|Q60854|SPB6_MOUSE .(TMT6plex)AFVEVNEEGTEAAAATAGMM(Oxidation)TVR
#> 2: sp|Q60854|SPB6_MOUSE .(TMT6plex)AFVEVNEEGTEAAAATAGMM(Oxidation)TVR
#> 3: sp|Q60854|SPB6_MOUSE .(TMT6plex)AFVEVNEEGTEAAAATAGMM(Oxidation)TVR
#> 4: sp|Q60854|SPB6_MOUSE .(TMT6plex)AFVEVNEEGTEAAAATAGMM(Oxidation)TVR
#> 5: sp|Q60854|SPB6_MOUSE .(TMT6plex)AFVEVNEEGTEAAAATAGMM(Oxidation)TVR
#> 6: sp|Q60854|SPB6_MOUSE .(TMT6plex)AFVEVNEEGTEAAAATAGMM(Oxidation)TVR
#> PrecursorCharge
#> 1: 3
#> 2: 3
#> 3: 3
#> 4: 3
#> 5: 3
#> 6: 3
#> PSM Condition
#> 1: .(TMT6plex)AFVEVNEEGTEAAAATAGMM(Oxidation)TVR_3_4359.56536443198 Long_HF
#> 2: .(TMT6plex)AFVEVNEEGTEAAAATAGMM(Oxidation)TVR_3_6190.04195694402 Long_HF
#> 3: .(TMT6plex)AFVEVNEEGTEAAAATAGMM(Oxidation)TVR_3_4359.56536443198 Long_HF
#> 4: .(TMT6plex)AFVEVNEEGTEAAAATAGMM(Oxidation)TVR_3_6190.04195694402 Long_HF
#> 5: .(TMT6plex)AFVEVNEEGTEAAAATAGMM(Oxidation)TVR_3_6190.04195694402 Long_HF
#> 6: .(TMT6plex)AFVEVNEEGTEAAAATAGMM(Oxidation)TVR_3_4359.56536443198 Long_HF
#> BioReplicate Run Channel Intensity Mixture TechRepMixture Fraction
#> 1: 21 3_3_3 1 NA 3 3_3 3
#> 2: 21 3_3_3 1 NA 3 3_3 3
#> 3: 24 3_3_3 4 NA 3 3_3 3
#> 4: 24 3_3_3 4 NA 3 3_3 3
#> 5: 26 3_3_3 6 NA 3 3_3 3
#> 6: 26 3_3_3 6 1820.072 3 3_3 3
If a user wants to use MSstatsConvert
package with data in a format that is not currently supported, it is enough to first re-format the data into a data.table
with column ProteinName, PeptideSequence, PrecursorCharge, FragmentIon, ProductCharge (with the latter two possibly equal to NA), Run and IsotopeLabelType (in case of non-TMT data) or Channel (in case of TMT data). Moreover, the dataset may include any column that will be used for filtering the dataset (for example a column that stores q-values). In our example, such additional columns are “Modifications” and “Score” from MaxQuant files.
Annotation columns should be called Condition and BioReplicate. For TMT data, Mixture, TechRepMixture columns may be added. Fractionation should be indicated by a Fraction column.
The goal of MSstatsPreprocess
function is to transform cleaned MS data into a format ready for statistical analysis with MSstats
or MSstatsTMT
packages. This function accepts several parameters, and each corresponds to a preprocessing step.
input
parameter is the dataset for preprocessing,annotation
is a description of biological conditions and replicates associated with MS runs (and channels in TMT case). If annotation is already included in the input
, it should be equal to NULL
. The annotation should be created by the MSstatsMakeAnnotation
function,feature_columns
is a vector of column names that will denote features,remove_shared_peptides
is a logical parameter - if TRUE
, shared peptides will be removed from the analysis. Currently, MSstats
assumes that only unique peptides are used, and presence of shared peptides may cause issues,remove_single_feature_proteins
is a logical parameter that indicates if proteins that only have a single feature should be removed from the analysis (TRUE
),feature_cleaning
is a list, that currently consists of two named elements: remove_features_with_few_measurements
should be equal to TRUE or FALSE. In the first case, feature that have less than three measurements across runs (or channels in a run for TMT data) will be removed. FALSE means that only features with no non-missing measurements will be removed. The summarize_multiple_psms
element should be a function that will be used to aggregate multiple feature measurements within a single MS run,aggregate_isotopic
is a logical parameter - TRUE
means that isotopic peaks will be aggregated (currently only used for Skyline input),columns_to_fill
is an optional named list with names corresponding to columns and values correponding to values that will be used for these columns. For example, if the dataset is missing information about ProductCharge
, such a column can be added by passing list(ProductCharge = NA)
to this parameter,score_filtering
, exact_filtering
and pattern_filtering
parameters are optional parameters that can be used to perform data filtering. An example is given below.maxquant_annotation = read.csv(system.file(
"tinytest/raw_data/MaxQuant/annotation.csv",
package = "MSstatsConvert"
))
maxquant_annotation = MSstatsMakeAnnotation(maxquant_cleaned,
maxquant_annotation,
Run = "Rawfile")
m_filter = list(col_name = "PeptideSequence",
pattern = "M",
filter = TRUE,
drop_column = FALSE)
oxidation_filter = list(col_name = "Modifications",
pattern = "Oxidation",
filter = TRUE,
drop_column = TRUE)
feature_columns = c("PeptideSequence", "PrecursorCharge")
maxquant_processed = MSstatsPreprocess(
maxquant_cleaned,
maxquant_annotation,
feature_columns,
remove_shared_peptides = TRUE,
remove_single_feature_proteins = FALSE,
pattern_filtering = list(oxidation = oxidation_filter,
m = m_filter),
feature_cleaning = list(remove_features_with_few_measurements = TRUE,
summarize_multiple_psms = max),
columns_to_fill = list("FragmentIon" = NA,
"ProductCharge" = NA,
"IsotopeLabelType" = "L"))
head(maxquant_processed)
#> Run PeptideSequence PrecursorCharge
#> 1: 121219_S_CCES_01_01_LysC_Try_1to10_Mixt_1_1 AEAPAAAPAAK 2
#> 2: 121219_S_CCES_01_02_LysC_Try_1to10_Mixt_1_2 AEAPAAAPAAK 2
#> 3: 121219_S_CCES_01_03_LysC_Try_1to10_Mixt_1_3 AEAPAAAPAAK 2
#> 4: 121219_S_CCES_01_05_LysC_Try_1to10_Mixt_2_2 AEAPAAAPAAK 2
#> 5: 121219_S_CCES_01_06_LysC_Try_1to10_Mixt_2_3 AEAPAAAPAAK 2
#> 6: 121219_S_CCES_01_08_LysC_Try_1to10_Mixt_3_2 AEAPAAAPAAK 2
#> Intensity ProteinName Condition BioReplicate Experiment IsotopeLabelType
#> 1: 4023100 P06959 1 1 1_1 L
#> 2: 5132500 P06959 1 1 1_2 L
#> 3: 2761600 P06959 1 1 1_3 L
#> 4: 4091800 P06959 2 2 2_2 L
#> 5: 4727000 P06959 2 2 2_3 L
#> 6: 2258400 P06959 3 3 3_2 L
#> FragmentIon ProductCharge
#> 1: NA NA
#> 2: NA NA
#> 3: NA NA
#> 4: NA NA
#> 5: NA NA
#> 6: NA NA
# OpenMS - TMT data
feature_columns_tmt = c("PeptideSequence", "PrecursorCharge")
openms_processed = MSstatsPreprocess(
openms_cleaned,
NULL,
feature_columns_tmt,
remove_shared_peptides = TRUE,
remove_single_feature_proteins = TRUE,
feature_cleaning = list(remove_features_with_few_measurements = TRUE,
summarize_multiple_psms = max)
)
head(openms_processed)
#> ProteinName PeptideSequence
#> 1: sp|Q60854|SPB6_MOUSE .(TMT6plex)AFVEVNEEGTEAAAATAGMM(Oxidation)TVR
#> 2: sp|Q60854|SPB6_MOUSE .(TMT6plex)AFVEVNEEGTEAAAATAGMM(Oxidation)TVR
#> 3: sp|Q60854|SPB6_MOUSE .(TMT6plex)AFVEVNEEGTEAAAATAGMM(Oxidation)TVR
#> 4: sp|Q60854|SPB6_MOUSE .(TMT6plex)AFVEVNEEGTEAAAATAGMM(Oxidation)TVR
#> 5: sp|Q60854|SPB6_MOUSE .(TMT6plex)AFVEVNEEGTEAAAATAGMM(Oxidation)TVR
#> 6: sp|Q60854|SPB6_MOUSE .(TMT6plex)AFVEVNEEGTEAAAATAGMM(Oxidation)TVR
#> PrecursorCharge PSM Condition
#> 1: 3 .(TMT6plex)AFVEVNEEGTEAAAATAGMM(Oxidation)TVR_3 Long_HF
#> 2: 3 .(TMT6plex)AFVEVNEEGTEAAAATAGMM(Oxidation)TVR_3 Long_HF
#> 3: 3 .(TMT6plex)AFVEVNEEGTEAAAATAGMM(Oxidation)TVR_3 Long_HF
#> 4: 3 .(TMT6plex)AFVEVNEEGTEAAAATAGMM(Oxidation)TVR_3 Long_HF
#> 5: 3 .(TMT6plex)AFVEVNEEGTEAAAATAGMM(Oxidation)TVR_3 Long_LF
#> 6: 3 .(TMT6plex)AFVEVNEEGTEAAAATAGMM(Oxidation)TVR_3 Long_M
#> BioReplicate Run Channel Intensity Mixture TechRepMixture Fraction
#> 1: 21 3_3_3 1 NA 3 3_3 3
#> 2: 24 3_3_3 4 NA 3 3_3 3
#> 3: 26 3_3_3 6 1820.0721 3 3_3 3
#> 4: 28 3_3_3 8 445.7412 3 3_3 3
#> 5: 25 3_3_3 5 1580.9510 3 3_3 3
#> 6: 23 3_3_3 3 1508.3302 3 3_3 3
#> FragmentIon
#> 1: <NA>
#> 2: <NA>
#> 3: <NA>
#> 4: <NA>
#> 5: <NA>
#> 6: <NA>
Annotation is created via the MSstatsMakeAnnotation
function. It takes the cleaned dataset and annotation file as input. Additionally, key-value pairs can be passed to this function to change column names (not including dots and other symbols) in the annotation from names given by values to names given by keys.
For programmatic applications and consistency of the interface, filtering is done with the help of lists.
For filtering based on numerical scores (for example q-value filtering), the list should consist of elements named
score_column
: name of a column that stores the score,score_threshold
: value above or below which measurements should be kept,direction
: if “greater”, values greater than score_threshold
will be kept; if “smaller”, values smaller than score_threshold
will be kept;behavior
: if “remove”, rows not below/above the threshold will be removed; if “replace”, intensity in rows not below/above the threshold will be replaced by a given value,handle_na
: if “keep”, NA
in the score column will not be removed,fill_value
: value that will be used if behavior = "replace"
,filter
: if TRUE
, filtering will be performed (can be used for conditional filtering),drop_column
: if TRUE
, column that stored the score will be removed.For example, to remove intensities smaller than 1, we could pass the following list to the score_filtering
parameters:
list(
list(score_column = "Intensity", score_threshold = 1,
direction = "greater", behavior = "remove",
handle_na = "remove", fill_value = NA, filter = TRUE, drop = FALSE
)
)
For filtering based on patterns (for example, removing oxidation peptides), the list should consist of elements named
col_name
: name of a column that values that may be removed,filter_symbols
: vector of values - rows with these values in col_name
will be removed or corresponding intensities will be replaced,behavior
: if “remove”, rows that contain filter_symbols
in col_name
will be removed; if “replace”, intensity in rows that contain filter_symbols
in col_name
will be replaced by a given value,fill_value
: value that will be used if behavior = "replace"
,filter
: if TRUE
, filtering will be performed (can be used for conditional filtering),drop_column
: if TRUE
, column that stored the score will be removed.For filtering based on exact values (for example, removing iRT proteins), the list should consists of elements named
col_name
: name of a column that stores strings that will be searched for given patterns,pattern
: vector of regular expressions - rows with matching values in col_name
will be removed,filter
: if TRUE
, filtering will be performed (can be used for conditional filtering),drop_column
: if TRUE
, column that stored the values for filtering will be removed.Finally, after preprocessing, MSstatsBalancedDesign
function can be applied to handle fractions and create balanced design. For label-free and SRM data, it means that fractionation or technical replicates will be detected if these information is not provided. Features measured in multiple fractions (overlapped) will be assigned to a unique fraction. Then, the data will be adjusted so that within each fraction, every feature has a row for certain run. If the intensity value is missing, it will be denoted by NA
.
For TMT data, a unique fraction will be selected for each overlapped feature and the data will adjusted so that within each run, every feature has a row for each channel. If the intensity is missing for a channel, it will be denoted by NA
.
maxquant_balanced = MSstatsBalancedDesign(maxquant_processed, feature_columns)
head(maxquant_balanced)
#> ProteinName PeptideSequence PrecursorCharge FragmentIon ProductCharge
#> 1 P06959 AEAPAAAPAAK 2 NA NA
#> 2 P06959 AEAPAAAPAAK 2 NA NA
#> 3 P06959 AEAPAAAPAAK 2 NA NA
#> 4 P06959 AEAPAAAPAAK 2 NA NA
#> 5 P06959 AEAPAAAPAAK 2 NA NA
#> 6 P06959 AEAPAAAPAAK 2 NA NA
#> IsotopeLabelType Condition BioReplicate
#> 1 L 1 1
#> 2 L 1 1
#> 3 L 1 1
#> 4 L 2 2
#> 5 L 2 2
#> 6 L 2 2
#> Run Fraction Intensity
#> 1 121219_S_CCES_01_01_LysC_Try_1to10_Mixt_1_1 1 4023100
#> 2 121219_S_CCES_01_02_LysC_Try_1to10_Mixt_1_2 1 5132500
#> 3 121219_S_CCES_01_03_LysC_Try_1to10_Mixt_1_3 1 2761600
#> 4 121219_S_CCES_01_04_LysC_Try_1to10_Mixt_2_1 1 2932900
#> 5 121219_S_CCES_01_05_LysC_Try_1to10_Mixt_2_2 1 4091800
#> 6 121219_S_CCES_01_06_LysC_Try_1to10_Mixt_2_3 1 4727000
dim(maxquant_balanced)
#> [1] 690 11
dim(maxquant_processed)
#> [1] 625 14
openms_balanced = MSstatsBalancedDesign(openms_processed, feature_columns_tmt)
head(openms_balanced)
#> ProteinName PeptideSequence
#> 1 sp|Q60854|SPB6_MOUSE .(TMT6plex)AFVEVNEEGTEAAAATAGMM(Oxidation)TVR
#> 2 sp|Q60854|SPB6_MOUSE .(TMT6plex)AFVEVNEEGTEAAAATAGMM(Oxidation)TVR
#> 3 sp|Q60854|SPB6_MOUSE .(TMT6plex)AFVEVNEEGTEAAAATAGMM(Oxidation)TVR
#> 4 sp|Q60854|SPB6_MOUSE .(TMT6plex)AFVEVNEEGTEAAAATAGMM(Oxidation)TVR
#> 5 sp|Q60854|SPB6_MOUSE .(TMT6plex)AFVEVNEEGTEAAAATAGMM(Oxidation)TVR
#> 6 sp|Q60854|SPB6_MOUSE .(TMT6plex)AFVEVNEEGTEAAAATAGMM(Oxidation)TVR
#> PrecursorCharge PSM Mixture
#> 1 3 .(TMT6plex)AFVEVNEEGTEAAAATAGMM(Oxidation)TVR_3 3
#> 2 3 .(TMT6plex)AFVEVNEEGTEAAAATAGMM(Oxidation)TVR_3 3
#> 3 3 .(TMT6plex)AFVEVNEEGTEAAAATAGMM(Oxidation)TVR_3 3
#> 4 3 .(TMT6plex)AFVEVNEEGTEAAAATAGMM(Oxidation)TVR_3 3
#> 5 3 .(TMT6plex)AFVEVNEEGTEAAAATAGMM(Oxidation)TVR_3 3
#> 6 3 .(TMT6plex)AFVEVNEEGTEAAAATAGMM(Oxidation)TVR_3 3
#> TechRepMixture Run Channel BioReplicate Condition Intensity
#> 1 3_3 3_3_3 1 21 Long_HF NA
#> 2 3_3 3_3_3 2 22 Norm 1068.580
#> 3 3_3 3_3_3 3 23 Long_M 1508.330
#> 4 3_3 3_3_3 4 24 Long_HF NA
#> 5 3_3 3_3_3 5 25 Long_LF 1580.951
#> 6 3_3 3_3_3 6 26 Long_HF 1820.072
dim(openms_balanced)
#> [1] 330 11
dim(openms_processed)
#> [1] 370 16
MSstatsBalancedDesign
output is a data.frame
of class MSstatsValidated
. Such a data.frame
will be recognized by statistical processing functions from MSstats
and MSstatsTMT
packages as a valid input, which will allow them to skip checks and transformation necessary to fit data into this format.