mzclust: EIC-clustering of LC-HRMS data.

Description

Based on the measurement partitions generated by mzagglom, extracted ion chromatograms (EICs) are assigned by a clustering procedure. Preparatory step for mzpick.

Usage

mzclust(MSlist,dmzdens=10,ppm=TRUE,drtdens=60,minpeak=4,maxint=1E6,
progbar=FALSE,merged=TRUE,from=FALSE,to=FALSE  )

Arguments

MSlist

MSlist returned by mzpart

dmzdens

Maximum measurement deviation (+/-) of m/z from its mean within each EIC

ppm

dmzdens given in ppm (TRUE) or as absolute value (FALSE)?

drtdens

Retention time (RT) tolerance for clustering; defined as (+/-) time units relative to the lowest and highest RT value in each cluster

minpeak

Minimum number of measurements expected in an EIC

maxint

EIC cluster with measurements above this intensity are kept, even if they do not fulfill minpeak

progbar

For debugging, ignore

merged

Merge EIC cluster of comparable m/z (TRUE or FALSE)?

from

For debugging, ignore

Value

Parameters: MSlist[[2]]: saves the parameter settings.
Scans: MSlist[[4]]: matrix with raw measurements and tags resorted for EIC clusters within the partition subsets.
EIC_index: MSlist[[6]]: Index assigning EIC cluster to sections in the raw measurement of MSlist[[4]]; required for fast (random) access.

Details

Within individual partitions calculated by mzagglom, an unsupervised clustering of measurements to individual ion chromatograms (EICs) is performed. For this purpose, a first EIC cluster is initialized with the most intense measurement, given an m/z uncertainty of 2*dmzdens. Along decreasing intensities, all other measurements are then sequentially either assigned to this cluster or used to define new clusters. For assignment, measurements must range both within the current tolerances of dmzdens and drtdens of an existing cluster. If several cluster are eligible for assignment, the one with the smallest mass difference between measurement m/z and cluster mean m/z will be used. Each time a new assignment to an existing cluster is made, its m/z estimate can be improved, i.e., the dmzdens tolerance around its mean m/z gradually shrinks from 2*dmzdens to dmzdens. In addition, dmzdens is used to update the RT tolerance of a cluster at each assignment.

With no measurements left, EIC clusters nested in m/z are then merged, relative to the m/z boundaries of the most intense cluster and stepwise along increasing mean m/z differences. Finally, EIC cluster are filtered to fulfill either minpeak or maxint.

References