addProcessing: Data manipulation and analysis methods
In rformassspectrometry/Spectra: Spectra Infrastructure for Mass Spectrometry Data
applyProcessing R Documentation
Data manipulation and analysis methods
Description
Various data analysis functions are available for Spectra objects. These
can be categorized into functions that either return a Spectra object
(with the manipulated data) and functions that directly return the
result from the calculation. For the former category, the data manipulations
are cached in the result object's processing queue and only exectuted
on-the-fly when the respective data gets extracted from the Spectra (see
section The processing queue for more information).
For the second category, the calculations are directly executed and the
result, usually one value per spectrum, returned. Generally, to reduce
memory demand, a chunk-wise processing of the data is performed.
Usage
applyProcessing(
object,
f = processingChunkFactor(object),
BPPARAM = bpparam(),
...
)
processingLog(x)
scalePeaks(x, by = sum, msLevel. = uniqueMsLevels(x))
## S4 method for signature 'Spectra'
addProcessing(object, FUN, ..., spectraVariables = character())
## S4 method for signature 'Spectra'
bin(
x,
binSize = 1L,
breaks = NULL,
msLevel. = uniqueMsLevels(x),
FUN = sum,
zero.rm = TRUE
)
## S4 method for signature 'Spectra'
containsMz(
object,
mz = numeric(),
tolerance = 0,
ppm = 20,
which = c("any", "all"),
BPPARAM = bpparam()
)
## S4 method for signature 'Spectra'
containsNeutralLoss(
object,
neutralLoss = 0,
tolerance = 0,
ppm = 20,
BPPARAM = bpparam()
)
## S4 method for signature 'Spectra'
entropy(object, normalized = TRUE)
## S4 method for signature 'ANY'
entropy(object, ...)
## S4 method for signature 'Spectra'
pickPeaks(
object,
halfWindowSize = 2L,
method = c("MAD", "SuperSmoother"),
snr = 0,
k = 0L,
descending = FALSE,
threshold = 0,
msLevel. = uniqueMsLevels(object),
...
)
## S4 method for signature 'Spectra'
replaceIntensitiesBelow(
object,
threshold = min,
value = 0,
msLevel. = uniqueMsLevels(object)
)
## S4 method for signature 'Spectra'
reset(object, ...)
## S4 method for signature 'Spectra'
smooth(
x,
halfWindowSize = 2L,
method = c("MovingAverage", "WeightedMovingAverage", "SavitzkyGolay"),
msLevel. = uniqueMsLevels(x),
...
)
## S4 method for signature 'Spectra'
spectrapply(
object,
FUN,
...,
chunkSize = integer(),
f = factor(),
BPPARAM = SerialParam()
)
Arguments
object
A Spectra object.
f
For spectrapply() and applyProcessing(): factor defining
how object should be splitted for eventual parallel processing.
Defaults to factor() for spectrapply() hence the object is not
splitted while it defaults to f = processingChunkSize(object) for
applyProcessing() splitting thus the object by default into chunks
depending on processingChunkSize().
BPPARAM
Parallel setup configuration. See BiocParallel::bpparam()
for more information. This is passed directly to the
backendInitialize() method of the MsBackend. See also
processingChunkSize() for additional information on parallel
processing.
...
Additional arguments passed to internal and downstream functions.
x
A Spectra.
by
For scalePeaks(): function to calculate a single numeric from
intensity values of a spectrum by which all intensities (of
that spectrum) should be divided by. The default by = sum will
divide intensities of each spectrum by the sum of intensities of that
spectrum.
msLevel.
integer defining the MS level(s) of the spectra to which
the function should be applied (defaults to all MS levels of object.
FUN
For addProcessing(): function to be applied to the peak matrix
of each spectrum in object.
For bin(): function to aggregate intensity values of peaks falling
into the same bin. Defaults to FUN = sum thus summing up intensities.
For spectrapply() and chunkapply(): function to be applied to
each individual or each chunk of Spectra.
spectraVariables
For addProcessing(): character with additional
spectra variables that should be passed along to the function defined
with FUN. See function description for details.
binSize
For bin(): numeric(1) defining the size for the m/z bins.
Defaults to binSize = 1.
breaks
For bin(): numeric defining the m/z breakpoints between
bins.
zero.rm
For bin(): logical(1) indicating whether to remove bins
with zero intensity. Defaults to TRUE, meaning the function will
discard bins created with an intensity of 0 to enhance memory
efficiency.
mz
For containsMz(): numeric with the m/z value(s) of the mass
peaks to check.
tolerance
For containsMz() and neutralLoss():
numeric(1) allowing to define a constant maximal accepted difference
between m/z values for peaks to be matched.
ppm
For containsMz() and neutralLoss(): numeric(1) defining a
relative, m/z-dependent, maximal accepted difference between m/z values
for peaks to be matched.
which
For containsMz(): either "any" or "all" defining whether
any (the default) or all provided mz have to be present in the
spectrum.
neutralLoss
for containsNeutralLoss(): numeric(1) defining the
value which should be subtracted from the spectrum's precursor m/z.
normalized
for entropy(): logical(1) whether the normalized
entropy should be calculated (default). See also
MsCoreUtils::nentropy() for details.
halfWindowSize
For pickPeaks(): integer(1), used in the
identification of the mass peaks: a local maximum has to be the
maximum in the window from (i - halfWindowSize):(i + halfWindowSize).
For smooth(): integer(1), used in the smoothing algorithm, the
window reaches from (i - halfWindowSize):(i + halfWindowSize).
method
For pickPeaks(): character(1), the noise estimators that
should be used, currently the the Median Absolute Deviation
(method = "MAD") and Friedman's Super Smoother
(method = "SuperSmoother") are supported.
For smooth(): character(1), the smoothing function that should be
used, currently, the Moving-Average- (method = "MovingAverage"),
Weighted-Moving-Average- (method = "WeightedMovingAverage"),
Savitzky-Golay-Smoothing (method = "SavitzkyGolay") are supported.
snr
For pickPeaks(): double(1) defining the
Signal-to-Noise-Ratio. The intensity of a local maximum has to be
higher than snr * noise to be considered as peak.
k
For pickPeaks(): integer(1), number of values left and right of
the peak that should be considered in the weighted mean calculation.
descending
For pickPeaks(): logical, if TRUE just values
betwee the nearest valleys around the peak centroids are used.
threshold
For pickPeaks(): a numeric(1) defining the proportion
of the maximal peak intensity. Only values above the threshold are
used for the weighted mean calculation.
For replaceIntensitiesBelow(): a numeric(1) defining the threshold
or a function to calculate the threshold for each spectrum on its
intensity values. Defaults to threshold = min.
value
For replaceIntensitiesBelow(): numeric(1) defining the
value with which intensities should be replaced with.
chunkSize
For spectrapply(): size of the chunks into which the
Spectra should be split. This parameter overrides parameters
f and BPPARAM.
Value
See the documentation of the individual functions for a description of the
return value.
Data analysis methods returning a Spectra
The methods listed here return a Spectra object as a result.
-
addProcessing(): adds an arbitrary function that should be applied to the
peaks matrix of every spectrum in object. The function (can be passed
with parameter FUN) is expected to take a peaks matrix as input and to
return a peaks matrix. A peaks matrix is a numeric matrix with two columns,
the first containing the m/z values of the peaks and the second the
corresponding intensities. The function has to have ... in its
definition. Additional arguments can be passed with .... With parameter
spectraVariables it is possible to define additional spectra variables
from object that should be passed to the function FUN. These will be
passed by their name (e.g. specifying spectraVariables = "precursorMz"
will pass the spectra's precursor m/z as a parameter named precursorMz
to the function. The only exception is the spectra's MS level, these will
be passed to the function as a parameter called spectrumMsLevel (i.e.
with spectraVariables = "msLevel" the MS levels of each spectrum will be
submitted to the function as a parameter called spectrumMsLevel).
Examples are provided in the package vignette.
-
bin(): aggregates individual spectra into discrete (m/z) bins. Binning is
performed only on spectra of the specified MS level(s) (parameter
msLevel, by default all MS levels of x). The bins can be defined with
parameter breaks which by default are equally sized bins, with size
being defined by parameter binSize, from the minimal to the maximal m/z
of all spectra (of MS level msLevel) within x. The same bins are used
for all spectra in x. All intensity values for peaks falling into the
same bin are aggregated using the function provided with parameter FUN
(defaults to FUN = sum, i.e. all intensities are summed up). Note that
the binning operation is applied to the peak data on-the-fly upon data
access and it is possible to revert the operation with the reset()
function (see description of reset() below).
-
countIdentifications: counts the number of identifications each scan has
led to. See countIdentifications() for more details.
-
pickPeaks(): picks peaks on individual spectra using a moving
window-based approach (window size = 2 * halfWindowSize). For noisy
spectra there are currently two different noise estimators available,
the Median Absolute Deviation (method = "MAD") and
Friedman's Super Smoother (method = "SuperSmoother"),
as implemented in the MsCoreUtils::noise().
The method supports also to optionally refine the m/z value of
the identified centroids by considering data points that belong (most
likely) to the same mass peak. Therefore the m/z value is calculated as an
intensity weighted average of the m/z values within the peak region.
The peak region is defined as the m/z values (and their respective
intensities) of the 2 * k closest signals to the centroid or the closest
valleys (descending = TRUE) in the 2 * k region. For the latter the k
has to be chosen general larger. See MsCoreUtils::refineCentroids() for
details.
If the ratio of the signal to the highest intensity of the peak is below
threshold it will be ignored for the weighted average.
-
replaceIntensitiesBelow(): replaces intensities below a specified
threshold with the provided value. Parameter threshold can be either
a single numeric value or a function which is applied to all non-NA
intensities of each spectrum to determine a threshold value for each
spectrum. The default is threshold = min which replaces all values
which are <= the minimum intensity in a spectrum with value (the
default for value is 0). Note that the function specified with
threshold is expected to have a parameter na.rm since na.rm = TRUE
will be passed to the function. If the spectrum is in profile mode,
ranges of successive non-0 peaks <= threshold are set to 0.
Parameter msLevel. allows to apply this to only spectra of certain MS
level(s).
-
scalePeaks(): scales intensities of peaks within each spectrum depending
on parameter by. With by = sum (the default) peak intensities are
divided by the sum of peak intensities within each spectrum. The sum of
intensities is thus 1 for each spectrum after scaling. Parameter
msLevel. allows to apply the scaling of spectra of a certain MS level.
By default (msLevel. = uniqueMsLevels(x)) intensities for all
spectra will be scaled.
-
smooth(): smooths individual spectra using a moving window-based approach
(window size = 2 * halfWindowSize). Currently, the
Moving-Average- (method = "MovingAverage"),
Weighted-Moving-Average- (method = "WeightedMovingAverage"),
weights depending on the distance of the center and calculated
1/2^(-halfWindowSize:halfWindowSize)) and
Savitzky-Golay-Smoothing (method = "SavitzkyGolay") are supported.
For details how to choose the correct halfWindowSize please see
MsCoreUtils::smooth().
Data analysis methods returning the result from the calculation
The functions listed in this section return immediately the result from the
calculation. To reduce memory demand (and allow parallel processing) the
calculations a chunk-wise processing is generally performed.
-
chunkapply(): apply an arbitrary function to chunks of spectra. See
chunkapply() for details and examples.
-
containsMz(): checks for each of the spectra whether they contain mass
peaks with an m/z equal to mz (given acceptable difference as defined by
parameters tolerance and ppm - see MsCoreUtils::common() for
details). Parameter
which allows to define whether any (which = "any", the default) or
all (which = "all") of the mz have to match. The function returns
NA if mz is of length 0 or is NA.
-
containsNeutralLoss(): checks for each spectrum in object if it has a
peak with an m/z value equal to its precursor m/z - neutralLoss (given
acceptable difference as defined by parameters tolerance and ppm).
Returns NA for MS1 spectra (or spectra without a precursor m/z).
-
entropy(): calculates the entropy of each spectra based on the metrics
suggested by Li et al. (https://doi.org/10.1038/s41592-021-01331-z).
See also MsCoreUtils::nentropy() in the MsCoreUtils package for
details.
-
estimatePrecursorIntensity(): defines the precursor intensities for MS2
spectra using the intensity of the matching MS1 peak from the
closest MS1 spectrum (i.e. the last MS1 spectrum measured before the
respective MS2 spectrum). With method = "interpolation" it is also
possible to calculate the precursor intensity based on an interpolation of
intensity values (and retention times) of the matching MS1 peaks from the
previous and next MS1 spectrum. See estimatePrecursorIntensity() for
examples and more details.
-
estimatePrecursorMz(): for DDA data: allows to estimate a fragment
spectra's precursor m/z based on the reported precursor m/z and the data
from the previous MS1 spectrum. See estimatePrecursorMz() for details.
-
neutralLoss(): calculates neutral loss spectra for fragment spectra. See
neutralLoss() for detailed documentation.
-
spectrapply(): applies a given function to each individual spectrum or
sets of a Spectra object. By default, the Spectra is split into
individual spectra (i.e. Spectra of length 1) and the function FUN
is applied to each of them. An alternative splitting can be defined with
parameter f. Parameters for FUN can be passed using ....
The returned result and its order depend on the function FUN and how
object is split (hence on f, if provided). Parallel processing is
supported and can be configured with parameter BPPARAM, is however only
suggested for computational intense FUN.
As an alternative to the (eventual parallel) processing of the full
Spectra, spectrapply() supports also a chunk-wise processing. For this,
parameter chunkSize needs to be specified. object is then split into
chunks of size chunkSize which are then (stepwise) processed by FUN.
This guarantees a lower memory demand (especially for on-disk backends)
since only the data for one chunk needs to be loaded into memory in each
iteration. Note that by specifying chunkSize, parameters f and
BPPARAM will be ignored.
See also chunkapply() above or examples below for details on chunk-wise
processing.
The processing queue
Operations that modify mass peak data, i.e. the m/z and intensity values of
a Spectra are generally not applied immediately to the data but are
cached within the object's processing queue. These operations are then
applied to the data only upon request, for example when m/z and/or
intensity values are extracted. This lazy execution guarantees that the
same functionality can be applied to any Spectra object, regardless of
the type of backend that is used. Thus, data manipulation operations can
also be applied to data that is read only. As a side effect, this enables
also to undo operations using the reset() function.
Functions related to the processing queue are:
-
applyProcessing(): for Spectra objects that use a writeable backend
only: apply all steps from the lazy processing queue to the peak data and
write it back to the data storage. Parameter f allows to specify how
object should be split for parallel processing. This should either be
equal to the dataStorage, or f = rep(1, length(object)) to disable
parallel processing alltogether. Other partitionings might result in
errors (especially if a MsBackendHdf5Peaks backend is used).
-
processingLog(): returns a character vector with the processing log
messages.
-
reset(): restores the data to its original state (as much as possible):
removes any processing steps from the lazy processing queue and calls
reset() on the backend which, depending on the backend, can also undo
e.g. data filtering operations. Note that a reset*( call after
applyProcessing() will not have any effect. See examples below for more
information.
Author(s)
Sebastian Gibb, Johannes Rainer, Laurent Gatto, Philippine Louail, Nir Shahaf, Mar Garcia-Aloy
See Also
-
compareSpectra() for calculation of spectra similarity scores.
-
processingChunkSize() for information on parallel and chunk-wise data
processing.
-
Spectra for a general description of the Spectra object.
Examples
## Load a `Spectra` object with LC-MS/MS data.
fl <- system.file("TripleTOF-SWATH", "PestMix1_DDA.mzML",
package = "msdata")
sps_dda <- Spectra(fl)
sps_dda
## -------- FUNCTIONS RETURNING A SPECTRA --------
## Replace peak intensities below 40 with a value of 1
sps_mod <- replaceIntensitiesBelow(sps_dda, threshold = 20, value = 1)
sps_mod
## Get the intensities of the first spectrum before and after the
## operation
intensity(sps_dda[1])
intensity(sps_mod[1])
## Remove all peaks with an intensity below 5.
sps_mod <- filterIntensity(sps_dda, intensity = c(5, Inf))
intensity(sps_mod)
## In addition it is possible to pass a function to `filterIntensity()`: in
## the example below we want to keep only peaks that have an intensity which
## is larger than one third of the maximal peak intensity in that spectrum.
keep_peaks <- function(x, prop = 3) {
x > max(x, na.rm = TRUE) / prop
}
sps_mod <- filterIntensity(sps_dda, intensity = keep_peaks)
intensity(sps_mod)
## We can also change the proportion by simply passing the `prop` parameter
## to the function. To keep only peaks that have an intensity which is
## larger than half of the maximum intensity:
sps_mod <- filterIntensity(sps_dda, intensity = keep_peaks, prop = 2)
intensity(sps_mod)
## With the `scalePeaks()` function we can alternatively scale the
## intensities of mass peaks per spectrum to relative intensities. This
## is specifically useful for fragment (MS2) spectra. We below thus
## scale the intensities per spectrum by the total sum of intensities
## (such that the sum of all intensities per spectrum is 1).
## Below we scale the intensities of all MS2 spectra in our data set.
sps_mod <- scalePeaks(sps_dda, msLevel = 2L)
## MS1 spectra were not affected
sps_mod |>
filterMsLevel(1L) |>
intensity()
## Intensities of MS2 spectra were scaled
sps_mod |>
filterMsLevel(2L) |>
intensity()
## Since data manipulation operations are by default not directly applied to
## the data but only cached in the internal processing queue, it is also
## possible to remove these data manipulations with the `reset()` function:
tmp <- reset(sps_mod)
tmp
lengths(sps_dda) |> head()
lengths(sps_mod) |> head()
lengths(tmp) |> head()
## Data manipulation operations cached in the processing queue can also be
## applied to the mass peaks data with the `applyProcessing()` function, if
## the `Spectra` uses a backend that supports that (i.e. allows replacing
## the mass peaks data). Below we first change the backend to a
## `MsBackendMemory()` and then use the `applyProcessing()` to modify the
## mass peaks data
sps_dda <- setBackend(sps_dda, MsBackendMemory())
sps_mod <- filterIntensity(sps_dda, intensity = c(5, Inf))
sps_mod <- applyProcessing(sps_mod)
sps_mod
## While we can't *undo* this filtering operation now using the `reset()`
## function, accessing the data would now be faster, because the operation
## does no longer to be applied to the original data before returning to the
## user.
## -------- FUNCTIONS RETURNING THE RESULT --------
## With the `spectrapply()` function it is possible to apply an
## arbitrary function to each spectrum in a Spectra.
## In the example below we calculate the mean intensity for each spectrum
## in a subset of the sciex_im data. Note that we can access all variables
## of each individual spectrum either with the `$` operator or the
## corresponding method.
res <- spectrapply(sps_dda[1:20], FUN = function(x) mean(x$intensity[[1]]))
head(res)
## As an alternative, applying a function `FUN` to a `Spectra` can be
## performed *chunk-wise*. The advantage of this is, that only the data for
## one chunk at a time needs to be loaded into memory reducing the memory
## demand. This type of processing can be performed by specifying the size
## of the chunks (i.e. number of spectra per chunk) with the `chunkSize`
## parameter
spectrapply(sps_dda[1:20], lengths, chunkSize = 5L)
## Precursor intensity estimation. Some manufacturers don't report the
## precursor intensity for MS2 spectra:
sps_dda |>
filterMsLevel(2L) |>
precursorIntensity()
## This intensity can however be estimated from the previously measured
## MS1 scan with the `estimatePrecursorIntensity()` function:
pi <- estimatePrecursorIntensity(sps_dda)
## This function returned the result as a `numeric` vector with one
## value per spectrum:
pi
## We can replace the precursor intensity values of the originating
## object:
sps_dda$precursorIntensity <- pi
sps_dda |>
filterMsLevel(2L) |>
precursorIntensity()
rformassspectrometry/Spectra documentation built on Dec. 19, 2024, 1:05 p.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
| applyProcessing | R Documentation |
Data manipulation and analysis methods
Description
Various data analysis functions are available for Spectra objects. These
can be categorized into functions that either return a Spectra object
(with the manipulated data) and functions that directly return the
result from the calculation. For the former category, the data manipulations
are cached in the result object's processing queue and only exectuted
on-the-fly when the respective data gets extracted from the Spectra (see
section The processing queue for more information).
For the second category, the calculations are directly executed and the result, usually one value per spectrum, returned. Generally, to reduce memory demand, a chunk-wise processing of the data is performed.
Usage
applyProcessing(
object,
f = processingChunkFactor(object),
BPPARAM = bpparam(),
...
)
processingLog(x)
scalePeaks(x, by = sum, msLevel. = uniqueMsLevels(x))
## S4 method for signature 'Spectra'
addProcessing(object, FUN, ..., spectraVariables = character())
## S4 method for signature 'Spectra'
bin(
x,
binSize = 1L,
breaks = NULL,
msLevel. = uniqueMsLevels(x),
FUN = sum,
zero.rm = TRUE
)
## S4 method for signature 'Spectra'
containsMz(
object,
mz = numeric(),
tolerance = 0,
ppm = 20,
which = c("any", "all"),
BPPARAM = bpparam()
)
## S4 method for signature 'Spectra'
containsNeutralLoss(
object,
neutralLoss = 0,
tolerance = 0,
ppm = 20,
BPPARAM = bpparam()
)
## S4 method for signature 'Spectra'
entropy(object, normalized = TRUE)
## S4 method for signature 'ANY'
entropy(object, ...)
## S4 method for signature 'Spectra'
pickPeaks(
object,
halfWindowSize = 2L,
method = c("MAD", "SuperSmoother"),
snr = 0,
k = 0L,
descending = FALSE,
threshold = 0,
msLevel. = uniqueMsLevels(object),
...
)
## S4 method for signature 'Spectra'
replaceIntensitiesBelow(
object,
threshold = min,
value = 0,
msLevel. = uniqueMsLevels(object)
)
## S4 method for signature 'Spectra'
reset(object, ...)
## S4 method for signature 'Spectra'
smooth(
x,
halfWindowSize = 2L,
method = c("MovingAverage", "WeightedMovingAverage", "SavitzkyGolay"),
msLevel. = uniqueMsLevels(x),
...
)
## S4 method for signature 'Spectra'
spectrapply(
object,
FUN,
...,
chunkSize = integer(),
f = factor(),
BPPARAM = SerialParam()
)
Arguments
object |
A |
f |
For |
BPPARAM |
Parallel setup configuration. See |
... |
Additional arguments passed to internal and downstream functions. |
x |
A |
by |
For |
msLevel. |
|
FUN |
For |
spectraVariables |
For |
binSize |
For |
breaks |
For |
zero.rm |
For |
mz |
For |
tolerance |
For |
ppm |
For |
which |
For |
neutralLoss |
for |
normalized |
for |
halfWindowSize |
For |
method |
For |
snr |
For |
k |
For |
descending |
For |
threshold |
For |
value |
For |
chunkSize |
For |
Value
See the documentation of the individual functions for a description of the return value.
Data analysis methods returning a Spectra
The methods listed here return a Spectra object as a result.
-
addProcessing(): adds an arbitrary function that should be applied to the peaks matrix of every spectrum inobject. The function (can be passed with parameterFUN) is expected to take a peaks matrix as input and to return a peaks matrix. A peaks matrix is a numeric matrix with two columns, the first containing the m/z values of the peaks and the second the corresponding intensities. The function has to have...in its definition. Additional arguments can be passed with.... With parameterspectraVariablesit is possible to define additional spectra variables fromobjectthat should be passed to the functionFUN. These will be passed by their name (e.g. specifyingspectraVariables = "precursorMz"will pass the spectra's precursor m/z as a parameter namedprecursorMzto the function. The only exception is the spectra's MS level, these will be passed to the function as a parameter calledspectrumMsLevel(i.e. withspectraVariables = "msLevel"the MS levels of each spectrum will be submitted to the function as a parameter calledspectrumMsLevel). Examples are provided in the package vignette. -
bin(): aggregates individual spectra into discrete (m/z) bins. Binning is performed only on spectra of the specified MS level(s) (parametermsLevel, by default all MS levels ofx). The bins can be defined with parameterbreakswhich by default are equally sized bins, with size being defined by parameterbinSize, from the minimal to the maximal m/z of all spectra (of MS levelmsLevel) withinx. The same bins are used for all spectra inx. All intensity values for peaks falling into the same bin are aggregated using the function provided with parameterFUN(defaults toFUN = sum, i.e. all intensities are summed up). Note that the binning operation is applied to the peak data on-the-fly upon data access and it is possible to revert the operation with thereset()function (see description ofreset()below). -
countIdentifications: counts the number of identifications each scan has led to. SeecountIdentifications()for more details. -
pickPeaks(): picks peaks on individual spectra using a moving window-based approach (window size =2 * halfWindowSize). For noisy spectra there are currently two different noise estimators available, the Median Absolute Deviation (method = "MAD") and Friedman's Super Smoother (method = "SuperSmoother"), as implemented in theMsCoreUtils::noise(). The method supports also to optionally refine the m/z value of the identified centroids by considering data points that belong (most likely) to the same mass peak. Therefore the m/z value is calculated as an intensity weighted average of the m/z values within the peak region. The peak region is defined as the m/z values (and their respective intensities) of the2 * kclosest signals to the centroid or the closest valleys (descending = TRUE) in the2 * kregion. For the latter thekhas to be chosen general larger. SeeMsCoreUtils::refineCentroids()for details. If the ratio of the signal to the highest intensity of the peak is belowthresholdit will be ignored for the weighted average. -
replaceIntensitiesBelow(): replaces intensities below a specified threshold with the providedvalue. Parameterthresholdcan be either a single numeric value or a function which is applied to all non-NAintensities of each spectrum to determine a threshold value for each spectrum. The default isthreshold = minwhich replaces all values which are <= the minimum intensity in a spectrum withvalue(the default forvalueis0). Note that the function specified withthresholdis expected to have a parameterna.rmsincena.rm = TRUEwill be passed to the function. If the spectrum is in profile mode, ranges of successive non-0 peaks <=thresholdare set to 0. ParametermsLevel.allows to apply this to only spectra of certain MS level(s). -
scalePeaks(): scales intensities of peaks within each spectrum depending on parameterby. Withby = sum(the default) peak intensities are divided by the sum of peak intensities within each spectrum. The sum of intensities is thus 1 for each spectrum after scaling. ParametermsLevel.allows to apply the scaling of spectra of a certain MS level. By default (msLevel. = uniqueMsLevels(x)) intensities for all spectra will be scaled. -
smooth(): smooths individual spectra using a moving window-based approach (window size =2 * halfWindowSize). Currently, the Moving-Average- (method = "MovingAverage"), Weighted-Moving-Average- (method = "WeightedMovingAverage"), weights depending on the distance of the center and calculated1/2^(-halfWindowSize:halfWindowSize)) and Savitzky-Golay-Smoothing (method = "SavitzkyGolay") are supported. For details how to choose the correcthalfWindowSizeplease seeMsCoreUtils::smooth().
Data analysis methods returning the result from the calculation
The functions listed in this section return immediately the result from the calculation. To reduce memory demand (and allow parallel processing) the calculations a chunk-wise processing is generally performed.
-
chunkapply(): apply an arbitrary function to chunks of spectra. Seechunkapply()for details and examples. -
containsMz(): checks for each of the spectra whether they contain mass peaks with an m/z equal tomz(given acceptable difference as defined by parameterstoleranceandppm- seeMsCoreUtils::common()for details). Parameterwhichallows to define whether any (which = "any", the default) or all (which = "all") of themzhave to match. The function returnsNAifmzis of length 0 or isNA. -
containsNeutralLoss(): checks for each spectrum inobjectif it has a peak with an m/z value equal to its precursor m/z -neutralLoss(given acceptable difference as defined by parameterstoleranceandppm). ReturnsNAfor MS1 spectra (or spectra without a precursor m/z). -
entropy(): calculates the entropy of each spectra based on the metrics suggested by Li et al. (https://doi.org/10.1038/s41592-021-01331-z). See alsoMsCoreUtils::nentropy()in the MsCoreUtils package for details. -
estimatePrecursorIntensity(): defines the precursor intensities for MS2 spectra using the intensity of the matching MS1 peak from the closest MS1 spectrum (i.e. the last MS1 spectrum measured before the respective MS2 spectrum). Withmethod = "interpolation"it is also possible to calculate the precursor intensity based on an interpolation of intensity values (and retention times) of the matching MS1 peaks from the previous and next MS1 spectrum. SeeestimatePrecursorIntensity()for examples and more details. -
estimatePrecursorMz(): for DDA data: allows to estimate a fragment spectra's precursor m/z based on the reported precursor m/z and the data from the previous MS1 spectrum. SeeestimatePrecursorMz()for details. -
neutralLoss(): calculates neutral loss spectra for fragment spectra. SeeneutralLoss()for detailed documentation. -
spectrapply(): applies a given function to each individual spectrum or sets of aSpectraobject. By default, theSpectrais split into individual spectra (i.e.Spectraof length 1) and the functionFUNis applied to each of them. An alternative splitting can be defined with parameterf. Parameters forFUNcan be passed using.... The returned result and its order depend on the functionFUNand howobjectis split (hence onf, if provided). Parallel processing is supported and can be configured with parameterBPPARAM, is however only suggested for computational intenseFUN. As an alternative to the (eventual parallel) processing of the fullSpectra,spectrapply()supports also a chunk-wise processing. For this, parameterchunkSizeneeds to be specified.objectis then split into chunks of sizechunkSizewhich are then (stepwise) processed byFUN. This guarantees a lower memory demand (especially for on-disk backends) since only the data for one chunk needs to be loaded into memory in each iteration. Note that by specifyingchunkSize, parametersfandBPPARAMwill be ignored. See alsochunkapply()above or examples below for details on chunk-wise processing.
The processing queue
Operations that modify mass peak data, i.e. the m/z and intensity values of
a Spectra are generally not applied immediately to the data but are
cached within the object's processing queue. These operations are then
applied to the data only upon request, for example when m/z and/or
intensity values are extracted. This lazy execution guarantees that the
same functionality can be applied to any Spectra object, regardless of
the type of backend that is used. Thus, data manipulation operations can
also be applied to data that is read only. As a side effect, this enables
also to undo operations using the reset() function.
Functions related to the processing queue are:
-
applyProcessing(): forSpectraobjects that use a writeable backend only: apply all steps from the lazy processing queue to the peak data and write it back to the data storage. Parameterfallows to specify howobjectshould be split for parallel processing. This should either be equal to thedataStorage, orf = rep(1, length(object))to disable parallel processing alltogether. Other partitionings might result in errors (especially if aMsBackendHdf5Peaksbackend is used). -
processingLog(): returns acharactervector with the processing log messages. -
reset(): restores the data to its original state (as much as possible): removes any processing steps from the lazy processing queue and callsreset()on the backend which, depending on the backend, can also undo e.g. data filtering operations. Note that areset*(call afterapplyProcessing()will not have any effect. See examples below for more information.
Author(s)
Sebastian Gibb, Johannes Rainer, Laurent Gatto, Philippine Louail, Nir Shahaf, Mar Garcia-Aloy
See Also
-
compareSpectra()for calculation of spectra similarity scores. -
processingChunkSize()for information on parallel and chunk-wise data processing. -
Spectra for a general description of the
Spectraobject.
Examples
## Load a `Spectra` object with LC-MS/MS data.
fl <- system.file("TripleTOF-SWATH", "PestMix1_DDA.mzML",
package = "msdata")
sps_dda <- Spectra(fl)
sps_dda
## -------- FUNCTIONS RETURNING A SPECTRA --------
## Replace peak intensities below 40 with a value of 1
sps_mod <- replaceIntensitiesBelow(sps_dda, threshold = 20, value = 1)
sps_mod
## Get the intensities of the first spectrum before and after the
## operation
intensity(sps_dda[1])
intensity(sps_mod[1])
## Remove all peaks with an intensity below 5.
sps_mod <- filterIntensity(sps_dda, intensity = c(5, Inf))
intensity(sps_mod)
## In addition it is possible to pass a function to `filterIntensity()`: in
## the example below we want to keep only peaks that have an intensity which
## is larger than one third of the maximal peak intensity in that spectrum.
keep_peaks <- function(x, prop = 3) {
x > max(x, na.rm = TRUE) / prop
}
sps_mod <- filterIntensity(sps_dda, intensity = keep_peaks)
intensity(sps_mod)
## We can also change the proportion by simply passing the `prop` parameter
## to the function. To keep only peaks that have an intensity which is
## larger than half of the maximum intensity:
sps_mod <- filterIntensity(sps_dda, intensity = keep_peaks, prop = 2)
intensity(sps_mod)
## With the `scalePeaks()` function we can alternatively scale the
## intensities of mass peaks per spectrum to relative intensities. This
## is specifically useful for fragment (MS2) spectra. We below thus
## scale the intensities per spectrum by the total sum of intensities
## (such that the sum of all intensities per spectrum is 1).
## Below we scale the intensities of all MS2 spectra in our data set.
sps_mod <- scalePeaks(sps_dda, msLevel = 2L)
## MS1 spectra were not affected
sps_mod |>
filterMsLevel(1L) |>
intensity()
## Intensities of MS2 spectra were scaled
sps_mod |>
filterMsLevel(2L) |>
intensity()
## Since data manipulation operations are by default not directly applied to
## the data but only cached in the internal processing queue, it is also
## possible to remove these data manipulations with the `reset()` function:
tmp <- reset(sps_mod)
tmp
lengths(sps_dda) |> head()
lengths(sps_mod) |> head()
lengths(tmp) |> head()
## Data manipulation operations cached in the processing queue can also be
## applied to the mass peaks data with the `applyProcessing()` function, if
## the `Spectra` uses a backend that supports that (i.e. allows replacing
## the mass peaks data). Below we first change the backend to a
## `MsBackendMemory()` and then use the `applyProcessing()` to modify the
## mass peaks data
sps_dda <- setBackend(sps_dda, MsBackendMemory())
sps_mod <- filterIntensity(sps_dda, intensity = c(5, Inf))
sps_mod <- applyProcessing(sps_mod)
sps_mod
## While we can't *undo* this filtering operation now using the `reset()`
## function, accessing the data would now be faster, because the operation
## does no longer to be applied to the original data before returning to the
## user.
## -------- FUNCTIONS RETURNING THE RESULT --------
## With the `spectrapply()` function it is possible to apply an
## arbitrary function to each spectrum in a Spectra.
## In the example below we calculate the mean intensity for each spectrum
## in a subset of the sciex_im data. Note that we can access all variables
## of each individual spectrum either with the `$` operator or the
## corresponding method.
res <- spectrapply(sps_dda[1:20], FUN = function(x) mean(x$intensity[[1]]))
head(res)
## As an alternative, applying a function `FUN` to a `Spectra` can be
## performed *chunk-wise*. The advantage of this is, that only the data for
## one chunk at a time needs to be loaded into memory reducing the memory
## demand. This type of processing can be performed by specifying the size
## of the chunks (i.e. number of spectra per chunk) with the `chunkSize`
## parameter
spectrapply(sps_dda[1:20], lengths, chunkSize = 5L)
## Precursor intensity estimation. Some manufacturers don't report the
## precursor intensity for MS2 spectra:
sps_dda |>
filterMsLevel(2L) |>
precursorIntensity()
## This intensity can however be estimated from the previously measured
## MS1 scan with the `estimatePrecursorIntensity()` function:
pi <- estimatePrecursorIntensity(sps_dda)
## This function returned the result as a `numeric` vector with one
## value per spectrum:
pi
## We can replace the precursor intensity values of the originating
## object:
sps_dda$precursorIntensity <- pi
sps_dda |>
filterMsLevel(2L) |>
precursorIntensity()
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.