metID: Combinatorial metabolite identification methods

In WMBEdmands/compMS2Miner: an automatable metabolite identification, visualization and data-sharing R package for high-resolution LC-MS datasets

Description

methods to facilitate metabolite identification including database monoisotopic mass matching, probable annotation filtration, mammalian Phase II metabolite prediction, molecular descriptor- random forest based retention time prediction, insilico metabolite fragmentation and nearest network neighbour metabolite chemical similarity scoring. In addition annotations can be automatically ranked and possible identities selected based on a mean consensus score based on mass accuracy, spectral database similarity, in silico fragmentation similarity, predicted retention time similarity, nearest network neighbour chemical similarity and crude plausibility ranking by PubMed repository text-mining. Optionally a divergent evolution approach to globally optimize the contributary weight of each consensus score using a training set of possible annotations contained in the "metID comments" table. The metaheuristic attempts to weight the include consensus scores according to how well the correct annotations are ranked amongst the possible best annotations.

Usage

metID(object, ...)

Arguments

...	option arguments to be passed along.
object.	a compMS2 class object
method.	method to use for metabolite identification. See details.

Details

Available methods:

1. monoisotopic mass annotation to data base resources (metID.dbAnnotate), currently available databases include HMDB, LMSD (lipidMaps) DrugBank, T3DB and ReSpect. possible metabolites electrospray adducts and substructure mass shifts are taken into account.

2. identifies most probable database annotations (metID.dbProb), taking into account substructure annotations identified by subStructure.Annotate.

3. Phase II metabolite identification from canonical SMILES currently available phase II metabolite prediction types include: acyl-, hydroxl- and amine- sulfates and glucuronides and glycine conjugates (metID.PredSMILES).

4. Retention time prediction using the molecular descriptors derived from the rcdk package and a randomForest recursive-feature elimination method of the caret package (metID.rtPred).

5. Combinatorial in silico fragment prediction using the command line version of MetFrag (metID.metFrag) or Competitive fragmentation modelling (CFM metID.CFM).

6. Correlation network from a peak table. This function calculates a correlation matrix from the peak areas/ height sample columns and creates a prefuse force directed correlation network that can then be visualized in the compMS2Explorer application. metID.corrNetwork

7. Spectral similarity network. Inter-spectrum spectral similarity scores (dot product) are calculated. Both fragment ion and precursor - fragment neutral loss pattern similarity scores are calculated and used to identify clusters of spectra with similar fragmentation/neutral loss patterns. A spectral similarity network is then calculated based on a minimum dot product score (minDotProdThresh, default = 0.8). The resulting network can then be visualized in the compMS2Explorer application. metID.specSimNetwork.

8. Correlation and spectral similarity based 1st Neighbour maximum chemical similarity scoring and optional automatic annotation identification (metID.chemSim).

9. build consensus annotations (metID.buildConsensus). This function seeks to rank annotation strength and automate metabolite identification based on 6 optional orthogonal annotation evidences, namely: mass accuracy, spectral database similarity (see function metID.matchSpectralDB), in silico fragmentation similarity (both metFrag and CFM see functions metID.metFrag and metID.CFM), random forest predicted retention time similarity (see function metID.rtPred), 1st network neighbour chemical similarity (both correlation and spectral similarity see function metID.chemSim) and finally crude literature based metabolite annotation strength by text-mining the PubMed repository using the Entrez system. The function can automatically add annotations to the 'metID comment" table of the compMS2Explorer application and also ranks the individual "best annotations" tables by the mean consensus metabolite annotation score.

10. optimize consensus annotations using the differential evolution algorithm of the DEoptim package.

Value

A compMS2 object with various metabolite identification information.

Source

Kenneth V. Price, Rainer M. Storn and Jouni A. Lampinen (2006). Differential Evolution - A Practical Approach to Global Optimization. Berlin Heidelberg: Springer-Verlag. ISBN 3540209506.

See Also

metID.dbAnnotate, metID.dbProb, metID.predSMILES, metID.reconSubStr, metID.metFrag, metID.chemSim, metID.corrNetwork, metID.specSimNetwork, metID.matchSpectralDB, metID.rtPred, metID.buildConsensus, metID.optimConsensus, metID.compMS2toMsp.

WMBEdmands/compMS2Miner documentation built on May 9, 2019, 10:04 p.m.