msAlign: Peak Alignment

Description Usage Arguments Details Value References See Also Examples

Description

Performs cross-spectral alignment of detected peaks.

Usage

1
msAlign(x, FUN="cluster", mz.precision=0.003, snr.thresh=10,...)

Arguments

x

An object of class msSet containing a "peak.list" element.

FUN

A character string specifying the method to use for alignment. Choices are

"cluster":

clusters peaks using one-dimensional hierarchical clustering and uses distances between peak locations as the similarity measure.

"gap":

analyzes peaks sequentially from low mass to high mass. Two adjacent peaks are classified into the same class if the distances between their locations is smaller than the specified threshold.

"vote":

clusters peaks iteratively. Each peak is associated with a window and the number of peaks that fall within the window across all samples is counted. The peak corresponding to the highest count forms a new peak cluster and all the peaks that have contributed to this peak are removed. The procedure is repeated until all peaks are exhausted from every sample.

"mrd":

clusters peaks by smoothing a histogram of scale-based feature locations for all spectra as identified by a call to msPeak(x,FUN="mrd", ...). The midpoints of the valleys in the smoothed histogram identifies the common peak locations across corresponding spectra.

Default: "cluster".

mz.precision

A numeric value, used to construct the threshold when performing clustering. The default value is 0.003 because SELDI data is often assumed to have +/- 0.3% mass drift, i.e., a peak at mass w could represent a protein with a mass within the interval [w(1-0.003), w(1+0.003)].

snr.thresh

A non-negative numeric value. The peaks with signal-to-noise ratio larger than this value will be used to construct the common set of peak classes. Default: 10.

...

Additional arguments passed to the msAlignMRD function.

Details

Currently, the mass accuracy of a mass spectrometer is proportional to the mass-to-charge (m/z) values. Thus, for a given set of spectra, the locations of the detected peaks will vary from spectrum to spectrum. In order to perform a comparative analysis of an ensemble of spectra, it is then a prerequisite to perform inter-sample alignment of the detected peaks. This process is normally called peak alignment or clustering.

The basic idea for peak alignment is to group peaks of similar molecular weight across all spectra into peak clusters or classes to form a superset, allowing for slight variations in mass. Each cluster is representative of a particular protein. Various methods have been proposed to align the peaks, which differ in how the superset is constructed.

Value

An object of class msSet, which is the input x with the added element "peak.class": a matrix with peak classes as rows and some summary statistics of the peak clusters as columns. These statistics include the location, left bound, right bound and peak span of the peak classes in both clock tick ("tick.loc", "tick.left", "tick.right", "tick.span") and mass measure ("mass.loc", "mass.left", "mass.right", "mass.span").

References

Coombes, K.R., Tsavachidis, S., Morris, J.S., Baggerly, K.A., and Kuerer, H.M., “Improved peak detection and quantification of mass spectrometry data acquired from surface-enhanced laser desorption and ionization by denoising spectra with the undecimated discrete wavelet transform," Proteomics, 5:4107–17, 2005.

Tibshirani, R., Hastie, T., Narasimhan, B., Soltys, S., Shi, G., Koong, A., and Le, Q.T., “Sample classification from protein mass spectrometry, by 'peak probability contrasts'," Bioinformatics, 20(17):3034–44, 2004.

Yasui, Y., McLerran, D., Adam, B.L., Winget, M., Thornquist, M., and Feng, Z., “An automated peak identification/calibration procedure for high-dimensional protein measures from mass spectrometers," Journal of Biomedicine and Biotechnology, 2003(4):242–8, 2003.

Yasui, Y., Pepe, M., Thompson, M.L., Adam, B.L., Wright, Jr., G.L., Qu, Y., Potter, J.D., Winget, M., Thornquist, M., and Feng, Z., “A data-analytic strategy for protein biomarker discovery: Profiling of high-dimensional proteomic data for cancer detection," Biostatistics, 4(3):449–63, 2003.

T.W. Randolph and Y. Yasui, Multiscale Processing of Mass Spectrometry Data, Biometrics, 62:589–97, 2006.

See Also

msPeak, msQuantify.

Examples

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
if (!exists("qcset")) data("qcset", package="msProcess")

## extract several spectra from the build-in
## dataset
z <- qcset[, 1:8]

## denoising
z <- msDenoise(z, FUN="wavelet", n.level=10, thresh.scale=2)

## local noise estimation
z <- msNoise(z, FUN="mean")

## baseline subtraction
z <- msDetrend(z, FUN="monotone", attach=TRUE)

## intensity normalization
z <- msNormalize(z)

## peak detection
z <- msPeak(z, FUN="simple", use.mean=FALSE, snr=2)

## peak alignment
z <- msAlign(z, FUN="cluster", snr.thresh=10,
    mz.precision=0.004)

## extract the peak.class
z[["peak.class"]]

## visualize the alignment
plot(z, process="msAlign", subset=1:8, offset=100,
    xlim=c(13000, 17000), lty=c(1,4))

zeehio/msProcess documentation built on May 4, 2019, 10:15 p.m.