In jasenfinch/MFassign: Molecular Formula Assignment For High Resolution ESI-MS Based Metabolomics Data

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.align = 'center'
)

Introduction

The assignments package provides an automated molecular formula assignment approach for ultra-high resolution electrospray ionisation mass spectrometry (ESI-MS) data from metabolomics experiments. This includes data from both direct and flow injection/infusion fingerprinting as well as liquid chromatograph mass spectrometry (LC-MS) profiling analytical techniques.

This vignette will provide a brief overview of the input data required, parameter selection, performing the assignments and assessing the results.

Before we begin, first load the package.

library(assignments)

Computational requirements and parallel processing

This approach is computationally intensive so the use of high-performance computing resources is recommended. A suggested minimum would be the use of 16 CPU workers and at least 8GB of RAM per worker (128GB total) to ensure that processing is completed in a reasonable duration.

The parallel back-end is provided by the future package. Information about the available parallel strategies can be found here. This example will use a relatively tiny data set so the following example parallel options will be used:

plan('multisession',workers = 2)

Input data

The requirements for data input are designed to be as simple as possible. Input data should consist of an m/z by sample intensity matrix with positive an negative mode data combined where available. The m/z features provided as column names should be in the form of <ionisation_mode><m/z>@<retention_time>. Ionisation mode should be given as a prefix n or p for negative or positive ionisation modes respectively. Feature m/z should be provided to an accuracy of least 5 decimal places. The retention time portion (@<retention_time>) is only required for LC-MS data and should be provided in minutes.

It is recommended that the data undergo pre-treatment routines such as relative standard deviation filtering, imputation and/or normalisation prior to assignment. However, this is not essential requirement and raw intensity values could also be used.

The input data for this example is a subset from an FIE-HRMS metabolomics experiment and is shown below. The feature intensities are total ion count (TIC) normalised.

feature_data

Parameters

Default parameters for a number of techniques are provided. The available techniques can be viewed as shown below.

availableTechniques()

The FIE-HRMS fingerprinting technique parameters would also be suitable for direct injection data. The default parameters are designed to be as widely applicable as possible and should suit many situations. For this example we will specify the use of the FIE-HRMS parameters.

parameters <- assignmentParameters('FIE-HRMS')

The parameters can then be viewed by printing the returned object.

parameters

It is possible to access and set all of these parameters. For example, the adducts method can be used to access the specified adducts:

adducts(parameters)

More accessor methods for assignment parameters can be viewed by running ?parameters.

Additional adducts, isotopes and transformations rules can also be appended to the relevant rules tables within the assignment parameters object. See the assignment parameters documentation for more information.

Assignment

To perform the molecular formula assignment we can execute the following.

assignment <- assignMFs(feature_data,
                        parameters)

For an overview of the assignment results, we can print the object.

assignment

Results

The following can be used to access the assigned m/z feature information.

assignments(assignment)

These feature assignments can also be summarised for each molecular formula.

summariseAssignments(assignment)

We can extract all the calculated correlations between the m/z features.

correlations(assignment)

As well as all the computed mathematical adduct, isotope and transformation relationships.

relationships(assignment)

To view all the iterations conducted during the assignment, the following can be used.

iterations(assignment)

Information for the component subgraphs identified in an iteration can also be extracted. These can either be the selected components or all the possible components.

components(assignment,
           iteration = 'A&I1',
           type = 'selected')

We can extract the tidygraph tbl_graph object for a given iterations.

graph(assignment,
      iteration = 'A&I1',
      type = 'selected')

Along with the graph of individual component.

component(assignment,
          component = 1,
          iteration = 'A&I1',
          type = 'selected')

And all the components of a given feature.

featureComponents(assignment,
                  feature = 'n191.01962',
                  type = 'all')

The graph of an individual component can be visualised. This first requires the ggraph package to be loaded.

library(ggraph)

plotComponent(assignment,
              component = 1,
              iteration = 'A&I1',
              type = 'selected')

We can also visualise the components containing a specific feature for a given iteration.

plotFeatureComponents(
  assignment,
  feature = 'n191.01962',
  iteration = 'A&I1',
  type = 'all',
  max_components = 6,
  axis_offset = 0.2
)

Because a molecular formula ranking threshold is applied during assignment, it may also be useful to generate all the alternative molecular formulas and their rankings for a specific m/z, adduct and isotope using the ipMF function from the mzAnnotation package.

mzAnnotation::ipMF(191.01962,
     adduct = '[M-H]1-',
     isotope = NA,
     ppm = 6)

jasenfinch/MFassign documentation built on Feb. 2, 2024, 11:21 a.m.