1 Foreword

pRolocGUI is under active development; current functionality is evolving and new features will be added. This software is free and open-source. You are invited to open issues in the Github pRolocGUI repository in case you have any questions, suggestions or have found any bugs or typos. To reach a broader audience for more general questions about proteomics analyses using R consider of writing to the Bioconductor Support Forum.

2 Introduction

This vignette describes the implemented functionality of the pRolocGUI package. The package is based on the MSnSet class definitions of MSnbase and on the functions defined in the pRoloc package. pRolocGUI is intended for, but not limited to, the interactive visualisation and analysis of quantitative spatial proteomics data. To achieve reactivity and interactivity, pRolocGUI relies on the shiny framework. We recommend some familiarity with the MSnSet class (see ?MSnSet for details) and the pRoloc vignette (see vignette("pRoloc-tutorial")) before using pRolocGUI.

There are 3 applications distributed with pRolocGUI which are wrapped and launched by the pRolocVis function. These 3 applications are called according to the argument app in the pRolocVis function which may be one of “pca”, “classify” or “compare”.

  • The pca application launches a Principal Components Analysis (PCA) plot of the data, with an alternate profiles tab for visualisation of protein profiles, it also features a searchable data table for the identification of proteins of interest.
  • The classify application has been designed to view machine learning classification results according to user-specified thresholds for the assignment of sub-cellular location.
  • The compare application allows the comparison of two comparable MSnSet instances, e.g. this might be of help for the analyses of changes in protein localisation in different conditions.

2.1 Getting started

Once R is started, the first step to enable functionality of the package is to load it, as shown in the code chunk below. We also load the pRolocdata data package, which contains quantitative proteomics datasets.

library("pRolocGUI")
library("pRolocdata")

We begin by loading the dataset hyperLOPIT2015 from the pRolocdata data package. The data was produced from using the hyperLOPIT technology on mouse E14TG2a embryonic stem cells (Christoforou et al 2016). For more background spatial proteomics data anlayses please see Gatto et al 2010, Gatto et al 2014 and also the pRoloc tutorial vignette.

data(hyperLOPIT2015)

To load one of the applications using the pRolocVis function and view the data you are required to specify a minimum of one key argument, object, which is the data to display and must be of class MSnSet (or a MSnSetList of length 2 for the compare application). Please see vignette("pRoloc-tutorial") or vignette("MSnbase-io") for importing and loading data. The argument app tells the pRolocVis function what type of application to load. One can choose from: "pca" (default), "classify", "compare". The optional argument fcol (and fcol1 and fcol2 for the compare app) can be used which allows the user to specify the feature meta-data label(s) (fData column name(s)) to be plotted. The default is markers (i.e. the labelled data) for the PCA and compare For the classification app one must specify the prediction column i.e. the feature meta-data label that corresponds to the column containing the classification results, generated from running a supervised machine learning analysis (see below).

For example, to load the default pRolocVis application:

pRolocVis(object = hyperLOPIT2015, fcol = "markers")

Launching any of the pRolocVis applications will open a new tab in a separate pop-up window, and then the application can be opened in your default Internet browser if desired, by clicking the ‘open in browser’ button in the top panel of the window.

To stop the applications from running press Esc or Ctrl-C in the console (or use the “STOP” button when using RStudio) and close the browser tab, where pRolocVis is running.

2.2 Which app should I use?

There are 3 different applications, each one designed to address a different specific user requirement.

  • The PCA app is intended for exploratory data analysis, which features a clickable interface and zoomable PCA plot. If you would like to search for a particular protein or set of proteins this is the application to use. This app also features a protein profiles tab, designed for examining the patterns of user-specified sets of proteins. For example, if one has several overlapping sub-cellular clusters in their data, as highlighted by the PCA plot or otherwise, one can check for separation in all data dimensions by examining the protein profile patterns. Proteins that co-localise are known to exhibit similar distributions (De Duve’s principale).
  • The classification app can be used for viewing the sub-cellular class predictions output from a supervised machine learning analysis and to help the user set a classification threshold (see the pRoloc tutorial for details on spatial proteomics data analysis).
  • The comparison application may be of interest if a user wishes to examine two replicate experiments, or two experiments from different conditions etc. Two PCA plots are loaded side-by-side and one can search and identify common proteins between the two data sets. As per the default application there is also a protein profiles tab to allow one to look at the patterns of protein profiles of interest in each dataset.

3 The pca application

The pca, default, application is characterised by an interactive and searchable Principal Components Analysis (PCA) plot. PCA is an ordinance method that can be used to transform a high-dimensional dataset into a smaller lower-dimenensional set of uncorrelated variables (principal components), such that the first principal component has the largest possible variance to account for as much variability in the data as possible. Each succeeding component in turn has the highest variance possible under the constraint that it be orthogonal to the preceding components. Thus, PCA is particularly useful for visualisation of multidimensional data in 2-dimensions, wherein all the proteins can be plotted on the same figure.

The application is subdivided in to three tabs: (1) PCA, (2) Profiles, and (3) Table Selection. A searchable data table containing the experimental feature meta-data is permanantly dispalyed at the bottom of the screen for ease. You can browse between the tabs by simply clicking on them at the top of the screen.

To run the pca application using pRolocVis:

pRolocVis(object = hyperLOPIT2015, fcol = "markers")
The PCA Tab

The PCA Tab

Viewing The PCA tab is characterised by its main panel which shows a PCA plot for the selected MSnSet. By default a PCA plot is used to display the data and the first two principal components are plotted. The sidebar panel controls what features to highlight on the PCA plot. Under the ‘Labels’ menu, input can be selected by clicking on and off the data class names, or by typing and searching in the white input box. Selected items can then be deleted, by clicking on the name of the class and pressing the delete button on your keyboard. The PCA plot will then be updated accordingly. Below the select box is a ‘transparancy’ slider bar which controls the opacity of the highlighted data classes and two action buttons ‘Zoom/reset plot’ and ‘Clear selection’, which are described below.