In fgcz/rawDiag: Brings Orbitrap Mass Spectrometry Data to Life; Fast and Colorful

BiocStyle::markdown()
knitr::opts_chunk$set(fig.wide = TRUE, fig.retina = 3, error=FALSE, eval=TRUE)

knitr::include_graphics("octopussy.png")

Introduction

Over the past two decades, liquid chromatography coupled to mass spectrometry (LC–MS) has evolved into the method of choice in the field of proteomics. [@Cox2011; @Mallick2010] During a typical LC–MS measurement, a complex mixture of analytes is separated by a liquid chromatography system coupled to a mass spectrometer (MS) through an ion source interface. This interface converts the analytes that elute from the chromatography system over time into a beam of ions. The MS records from this ion beam a series of mass spectra containing detailed information on the analyzed sample. [@Savaryn2016] The resulting raw data consist of the mass spectra and their metadata, typically recorded in a vendor-specific binary format. During a measurement the mass spectrometer applies internal heuristics, which enables the instrument to adapt to sample properties, for example, sample complexity and amount of ions in near real time. Still, method parameters controlling these heuristics need to be set prior to the measurement. Optimal measurement results require a careful balancing of instrument parameters, but their complex interactions with each other make LC–MS method optimization a challenging task.

Here we present r BiocStyle::Biocpkg('rawDiag'), a platform-independent software tool implemented in the R language [@newS] that supports LC–MS operators during the process of empirical method optimization. Our work builds on the ideas of the discontinued software rawMeat (VAST Scientific). Our application is currently tailored toward spectral data acquired on Thermo Fisher Scientific instruments (raw format), with a particular focus on Orbitrap [@Zubarev2013] mass analyzers (Exactive or Fusion instruments). These instruments are heavily used in the field of bottom-up proteomics [@Aebersold2003] to analyze complex peptide mixtures derived from enzymatic digests of proteomes.

r BiocStyle::Biocpkg('rawDiag') is meant to run after MS acquisition, optimally as an interactive R shiny application, and produces a series of diagnostic plots visualizing the impact of method parameter choices on the acquired data across injections. If static reports are required then pdf files can be generated using r BiocStyle::CRANpkg('rmarkdown'). In this vignette, we present the usage of our tool.

r BiocStyle::Biocpkg('rawDiag') gains advantages from being part of the Bioconductor ecosystem, such as its ability to utilize the r BiocStyle::Biocpkg('rawrr') package and potentially extend its functionality through interaction with the r BiocStyle::Biocpkg('Spectra') infrastructure, particularly with the r BiocStyle::Biocpkg('MsBackendRawFileReader').

Requirements

r BiocStyle::Biocpkg('rawDiag') proides a wrapper function readRaw using the r BiocStyle::Biocpkg('rawrr') methods raw::readIndex, rawrr::readTrailer, and rawrr::readChromatogram to read proprietary mass spectrometer generated data by invoking third-party managed methods through a system2 text connection. The r BiocStyle::Biocpkg('rawrr') package provides the entire stack below, which r BiocStyle::Biocpkg('rawDiag') utilizes.

`R>`

`text connection`

`system2`

Mono Runtime

Managed Assembly (CIL/.NET code) rawrr.exe

ThermoFisher.CommonCore.*.dll

Install

OS dependencies

Linux (debian/ubuntu)

In case you prefer to compile rawrr.exe from C# source code, please install the mono compiler and xbuild by installing the following Linux packages:

sudo apt-get install mono-mcs mono-xbuild

Otherwise, to execute the precompiled code, the following Linux packages are sufficient:

sudo apt-get install mono-runtime libmono-system-data4.0-cil -y

macOS (Catalina/BigSur)

brew install mono

or install from

https://www.mono-project.com/

Microsoft Windows

Running the rawrr.exe will run out of the box.

If the native C# compiler is not available install mono from:

https://www.mono-project.com/

R

To install this package, start R (version ">=4.4") and enter:

if (!require("BiocManager", quietly = TRUE))
    install.packages("BiocManager")

BiocManager::install("rawrr")

assemblies aka Common Intermediate Language bytecode - the download and install can be done on all platforms using the command:

rawDiag::checkRawrr
rawDiag::checkRawrr()

if (isFALSE(rawrr::.checkDllInMonoPath())){
  rawrr::installRawFileReaderDLLs()
}
rawrr::installRawrrExe()

for more information please read the INSTALL file in the r BiocStyle::Biocpkg('rawrr') package.

Usage

R command line

Input

fetch example Orbitrap raw files from r BiocStyle::Biocpkg('ExperimentHub')'s r BiocStyle::Biocpkg('tartare') package.

library(ExperimentHub)
ExperimentHub::ExperimentHub() -> eh
normalizePath(eh[["EH3222"]]) -> EH3222
normalizePath(eh[["EH4547"]]) -> EH4547

(rawfileEH3222 <- paste0(EH3222, ".raw"))
if (!file.exists(rawfileEH3222)){
  file.copy(EH3222, rawfileEH3222)
}

(rawfileEH4547 <- paste0(EH4547, ".raw"))
if (!file.exists(rawfileEH4547)){
  file.copy(EH4547, rawfileEH4547)
}

c(rawfileEH3222, rawfileEH4547) -> rawfile

Of note, the proprietary .Net assemblies [@RFR] require a file extentention of .raw. Therfore we have to rename the EH files and add the .raw suffix.

list meta data of the raw files.

(rawfile |>
  lapply(FUN = rawrr::readFileHeader) -> rawFileHeader)

readRaw - read Orbitrap raw file

read the two instrument raw files by using the r BiocStyle::Biocpkg('rawDiag') package.

rawfile |>
  BiocParallel::bplapply(FUN = rawDiag::readRaw) |>
  Reduce(f = rbind) -> x

Output - Visualization

This package provides several plot functions tailored toward MS data. The following list shows all available plot methods.

library(rawDiag)
ls("package:rawDiag") |>
  grep(pattern = '^plot', value = TRUE) -> pm

pm |>
  knitr::kable(col.names = "package:rawDiag plot functions")

An inherent problem of visualizing data is the fact that depending on the data at hand, specific visualizations lose their usefulness, e.g., overplotting in a scatter plot if too many data points are present. To address this problem, we implemented most of the plot functions in different versions inspired by the work of @Cleveland1993, @Sarkar2008, and @Wickham2009. The data can be displayed in trellis plot manner using the faceting functionality of r BiocStyle::CRANpkg('ggplot2'). Alternatively, overplotting using color coding or violin plots based on descriptive statistics values can be chosen, which allows the user to interactively change the appearance of the plots based on the situation at hand. For instance, a large number of files are best visualized by violin plots, giving the user an idea about the distribution of the data points. On the basis of this, a smaller subset of files can be selected and visualized with another technique.

The code snippet below applies all plot methods on the example data.

pm |> 
  lapply(FUN = function(plotFUN) {
    lapply(c('trellis'), function(method) {
      message("plotting", plotFUN, "using method", method, "...")
      do.call(plotFUN, list(x, method)) 
    })
  }) 

The appearance of each plot depends on the instrument, sample, and method used to acquire the data. Therefore, it is hard to say what each ideal plot should look like. In particular, in the example above, we use data generated on an r rawFileHeader[[1]]$"Instrument name", r rawFileHeader[[1]]$"RAW file" and r rawFileHeader[[2]]$"Instrument name", r rawFileHeader[[2]]$"RAW file" instrument using data-independent acquisition (DIA) [@Bruderer2017] and data-dependent acquisition (DDA) methods. For more information on the plot methods and its application, please read the package man pages and the application examples in the manuscript [@Trachsel2018].

Launching the shiny application

The package provides a simple interactive r BiocStyle::Biocpkg('shiny')-based graphical user interface for exploring Thermo Fisher Scientific raw data.

If you have a directory containing raw files, you can create a shiny application as follows:

rawfile |>
  dirname() |>
  rawDiag::buildRawDiagShinyApp() -> app

The r BiocStyle::Biocpkg('shiny') runApp function launches the app in our browser.

shiny::runApp(app)

By default, the application lets you choose the raw files in the provided directory and provides the visualizations of the raw data as output. The user can interactively change the by the r BiocStyle::Biocpkg('rawDiag') the package provided plot functions and arguments.

Additionally, the application provides PDF generation and download buttons. Optionally height and width can be changed in the user interface.

Of note, the rawDiag::rawDiagServer module can be integrated into an existing r BiocStyle::CRANpkg('shinydashboard') application, e.g., https://shiny-ms.fgcz.uzh.ch/fgczmsqc-dashboard/.

FAQ

I would like to load multiple files into a single data.frame to do comparisons; what is the preferred method for doing so?

consider all raw files of your working directory, e.g., ~/Downloads and load them.

file.path(Sys.getenv("HOME"), "Downloads") |>
  setwd()

list.files() |>
  grep(pattern = '*.raw$', value = TRUE) |> 
  lapply(FUN = rawDiag::readRaw) |> 
  Reduce(f = rbind) -> x

as alternative to lapply you can utilize the r BiocStyle::Biocpkg('BiocParallel') package bplapply function.

References {-}

Session information {-}

sessionInfo()

fgcz/rawDiag documentation built on May 5, 2024, 4:23 a.m.