pRolocGUI: Interactive visualisation of organelle (spatial) proteomics data

Thomas Naake and Laurent Gatto*

*Computational Proteomics Unit, University of Cambridge

Foreword

This vignette describes the implemented functionality in the pRolocGUI package. The package is based on the MSnSet class definitions of MSnbase (Gatto and Lilley, 2012) and on the functions defined in the package pRoloc (Breckels, Gatto, Christoforou, Groen, et al., 2013; Gatto, Breckels, Wieczorek, Burger, et al., 2014). pRolocGUI is intended for the visualisation and analysis of proteomics data, especially for the analyses of LOPIT (Dunkley, Hester, Shadforth, Runions, et al., 2006) or PCP (Foster, Hoog, Zhang, Zhang, et al., 2006) experiments.
To achieve reactivity and interactivity, pRolocGUI relies on the shiny framework.

The implemented application facilitates a higher degree of interactivity with the underlying spatial proteomics data: The distributed functions pRolocVis and pRolocComp offer interactive Principal Component Analysis (PCA) plots and protein profile plots, as well as exploration of quantitative and qualitative meta-data. Key features of pRolocVis and pRolocComp are the identification of features in plots, a 'reverse search' based on querying meta-data which allows for highlighting the features on plots and an import/export functionality by using the FeaturesOfInterest/FoICollection infrastructure distributed by the MSnbase package. Additionally, pRolocComp allows for comparison of two comparable MSnSet instances, e.g. this might be of great help for analyses of changes in protein localisation in different MSnSets.

We recommend some familiarity with the MSnSet class (see ?MSnSet for details) and the pRoloc vignette (available with vignette("pRoloc-tutorial")).

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 contact Laurent Gatto (lg390@cam.ac.uk) or Thomas Naake (naake@stud.uni-heidelberg.de) 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 list.


  1. Introduction

Currently, pRolocGUI distributes the pRolocVis and pRolocCompapplication.

The function pRolocVis needs an object of class MSnSet or a list of MSnSet objects as an argument, while pRolocComp needs a list containing two instances of class MSnSet. To prepare the environment to run a pRolocVis/pRolocComp session, the pRolocGUI package and for demonstration purposes four example MSnSets are loaded to the environment. The example data sets are available from the pRolocdata (Gatto and Breckels, 2014) experiment package and are derived from experiments andy2011 from Breckels, Gatto, Christoforou, Groen, et al. (2013), tan2009r1 and tan2009r2, the first and second replicate from Tan, Dvinge, Christoforou, Bertone, et al. (2009) and dunkley2006 from Dunkley, Hester, Shadforth, Runions, et al. (2006).

library("pRolocGUI")
data(andy2011, package = "pRolocdata")
data(tan2009r1, package = "pRolocdata")
data(tan2009r2, package = "pRolocdata")
data(dunkley2006, package = "pRolocdata")
pRolocVis needs an object of class MSnSet as an argument. We can launch the application with an MSnSet by assigning it to the argument object:
pRolocVis(object = andy2011)
Alternatively, to upload multiple objects of class MSnSet, pRolocVis accepts both lists with named and unnamed objects. This allows for analysis of multiple data sets without stopping the application from running. The names of objects of lists will appear in the drop-down menu in the data tab, while lists with unnamed objects will have a drop-down menu with automatically named entries as a consequence, i.e. object 1 ... object n, where n is the length of the list.
namedVis <- list(andy2011 = andy2011, tan2009r1 = tan2009r1, dunkley2006 = dunkley2006)
unnamedVis <- list(andy2011, tan2009r1, dunkley2006)
pRolocVis(object = namedVis)
pRolocVis(object = unnamedVis)
pRolocComp requires a list of two MSnSets which can be named or unnamed.
namedComp <- list(tan2009r1 = tan2009r1, tan2009r2 = tan2009r2)
unnamedComp <- list(tan2009r1, tan2009r2)
pRolocComp(object = namedComp)
pRolocComp(object = unnamedComp)

N.B.! It is also possible to use a partly named list in pRolocVis and pRolocComp.

Launching pRolocVis or pRolocComp will open a new tab in your default Internet browser.

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/pRolocComp is running.


  1. Tabs of pRolocVis and pRolocComp

To optimise ease of use the interfaces of pRolocVis and pRolocComp are subdivided in seven tabs:

  • PCA,
  • protein profiles,
  • quantitation,
  • feature meta-data,
  • sample meta-data and
  • search,
  • data.

You browse through the tabs by simply clicking on them. Each tab selected will have a different kind of appearance while some (PCA, protein profiles, quantitation and feature meta-data) share a common feature in the sidebar, the Display selection widget (see section 3. Display selection widget for further details).

In case you have a question and want to consult the vignette for a certain issue (e.g. regarding PCA tab or on how to use the Display selection widget) click on ? which will open the vignette in a new browser tab in the corresponding section.

In general, the tabs for pRolocVis and pRolocComp will look alike. The tab data however differs for the two applications. While in pRolocVis this tab allows for the selection of the used MSnSet and the upload of .Rda or .Rdata files, in pRolocComp there is the possibility to subset the used MSnSets (in terms of using common, unique and common & unique features in the two MSnSets used) as well as to submit features for selection. See 2.7. data for further details.

Fig. 1: Vignette of pRolocVis and pRolocComp

2.1. PCA

The tab PCA is characterised by its main panel which shows a PCA plot for the selected MSnSet in the case of pRolocVis and two PCA plots for pRolocVis. The sidebar panel is divided into Display selection and Plot.

Fig. 2: Appearance of PCA tab for pRolocVis (andy2011)

For pRolocComp the plot whose appearance is going to altered has to be selected by selecting the appropriate radio button left of the name of the MSnSet instance in the sidebar panel.

In addition, pRolocComp offers the possibility to mirror the PCA plot of the second object. By clicking on x-axis below mirror 2nd object features are mirrored along the x-axis, while clicking on y-axis mirrors along the y-axis. This may be important when you compare experiments whose PCA analysis have different signs.

Fig. 3: Radio buttons to select MSnSet in pRolocComp. Argument object is a named list with MSnSet instances tan2009r1 and tan2009r2

The manipulation of the plots works the same way for pRolocVis and for pRolocComp:

The Display selection widget is described below. The Plot compartment enables to adjust the appearance of the PCA plot in the main panel. We are able to colour features (proteins) in matters of common properties by changing the drop-down list colour. These properties are the MSnSet's feature variables. For example if we upload the andy2011 data set in pRolocVis, and select the colour markers, the features in the PCA plot will be coloured according to their organelle affiliation. As soon as we select another colour than none, two (or three) new items will be added to the Plot widget:

  1. symbol type: By selecting one of the feature variables of the MSnSet in the drop-down list of symbol type the symbol type of the features in the plot will be changed.

  2. legend and position of legend: By clicking on the check box to the left of legend a legend is added to the plot and by choosing one of the items in the drop-down list position of legend below its position will be changed.

  3. point size: This drop-down list might appear when numeric feature variable have been identified. The default 1 allows for an unaltered display of the plot, while selecting other items in the list renders the features in the PCA plot according to their numerical value in the variable label (for example classification scores).

Fig. 4: Appearance of PCA tab (andy2011). markers used for colours, legend added.

By changing the drop-down lists of the items PC along x axis and PC along y axis the x-values and y-values, respectively, the plot will be rendered according to the new principal components.

To zoom in and out drag and drop the little arrows of the slider of the items zoom x-axis and zoom y-axis. This may be of great help when you want to identify points in dense clusters.

By clicking on Download Plot in the main panel below the PCA plot will open a dialog window with an interface on showing or saving the PCA plot as it is displayed in the main panel.

2.2. protein profiles

For pRolocVis the tab protein profiles shows the protein profiles in the main panel (with an option of exporting the plot as it is shown in the main panel by clicking on the button Download Plot) - for pRolocComp it shows the plots for the two MSnSet instances (the corresponding plot(s) for the first element in the list will be displayed on the left, for the second element on the right). In the sidebar panel there is the Display selection widget and the Plot widget.

Have a look on section 3. Display selection widget if you want to retrieve information about how to use the Display selection widget.

The Plot widget helps to manipulate the plots shown in the main panel (in pRolocComp one has first to select the appropriate radio button next to the MSnSet instance in the sidebar panel). Let's assume we want to have a look upon the protein profiles for the proteins from which we know that they belong to the organelles Endoplasmic reticulum, the Golgi apparatus, Mitochondrion and the plasma membrane for the andy2011 MSnSet. This is done in pRolocVis, but works in the same way in pRolocComp. We have four organelles to look at, so we select 4 (in pRolocComp there is either the possibility to select 1 or 2 plots per MSnSet) in the drop-down list number of plots to display. We will select the feature variable markers in the drop-down list feature(s) in and select ER (coding for Endoplasmic reticulum) in the drop-down list underneath (assigned to). To display the next plot we have to change the slider Selected plot to position 2. Accordingly to our question we will change the second drop-down list to Golgi (coding for Golgi apparatus). We proceed with the two remaining organelles as described before by changing firstly the slider to the next position and by changing the drop-down lists accordingly to the organelles we want to display. Please be aware that it is possible to "go" back to a plot to change its parameter.

Fig. 5: Appearance of protein profiles tab in pRolocVis showing protein profiles of organelles/compartments Endoplasmic reticulum, Golgi apparatus, Mitochondrion and plasma membrane of markers (andy2011)

2.3. quantitation

The tab quantitation displays the quantitation data for the proteins as a data table.

In the main panel you can change the number of proteins shown per page and search both for proteins (or for the quantitation data). Also, you may sort the proteins by name or the quantitation data by clicking on the arrows on the top of the data table.

In the sidebar panel the Display selection widget is located as well as radio buttons to display all data or just selected features (see 3. Display selection for further details). In pRolocComp there is also another well panel to select the appropriate radio button next to the name of the MSnSet to show the respective quantitation data.

Fig. 6: Appearance of quantitation tab (andy2011) in pRolocVis. Features shown originate from selection made in the PCA and protein profiles plots

2.4. feature meta-data

The tab feature meta-data displays the feature meta-data for the proteins as a data table.

The layout of the tab is similar to the quantitation tab and allows for sorting and querying the feature meta-data of the selected MSnSet.

The sidebar comprises the Display selection widget and radio buttons to show all or only selected features (see 3. Display selection for further details). In pRolocComp there is in addition a set of two radio buttons which allow to switch the MSnSet shown, thus, the feature meta-data will be rendered to the selected MSnSet.

Fig. 7: Appearance of feature meta-data tab (andy2011) in pRolocVis. Features shown originate from selection made in protein profiles plot

2.5. sample meta-data

The tab sample meta-data displays the sample meta-data for the experiment, the name of the isotopes used for tagging and the associated fractions.

In pRolocComp select the appropriate radio button next to the object name in sidebar panel to display the corresponding sample meta-data.

Fig. 8: Appearance of sample meta-data tab (andy2011) in pRolocVis

The appearance and operation are identical for pRolocVis and pRolocGUI.

pRolocVis and pRolocComp allow to use past search results to display in the PCA plot, protein profiles and in the tabs quantitation and feature meta-data (see 3. Display selection for further details if this is your intention). This ability requires the object pRolocGUI_SearchResults in the global environment which is of class FeaturesOfInterest or FoICollection (enter ?FeaturesOfInterest in the console for further details).

In case this objects exists it will automatically be loaded to pRolocVis/pRolocComp and its content is displayed in the tab search. Use the drop-down list in the main panel to browse through the different features of the FoICollection. To select features and display them in multiple tabs add them to the field in the multiple drow-down list.

If no object called pRolocGUI_SearchResults exists in the global environment you still have the possibility to assign FeaturesOfInterest to an FoICollection which will be assigned to the global environment when exiting pRolocVis/pRolocComp.

To save features of interest to the object internally you need to select features and add these to the FoICollection by entering an appropriate description in the text field (on the sidebar panel, which will be useful to trace back to the underlying features and does not exist yet in the FoICollection). Add the selected features to the object pRolocGUI_SearchResults in the global environment by clicking on Create new features of interest. You only have the possibility to add selected features to the FoICollection when you have entered an appropriate description, i.e. one that doesn't exist yet in the FoICollection and if you have selected FeaturesOfInterest, otherwise the button does not show up in the application.

When exiting pRolocVis/pRolocComp the FoICollection will be assigned to the object pRolocGUI_SearchResults in the global environment.

To create an example object pRolocGUI_SearchResults containing the first ten features of tan2009r1 run the following commands in the console. Both traceable and non-traceable FeaturesOfInterest/FoICollection are usable by pRolocVis/pRolocComp.

data(tan2009r1, package = "pRolocdata")
pRolocGUI_SearchResults <- FoICollection()
newFeat <- FeaturesOfInterest(description = "test_01",
                              fnames = featureNames(tan2009r1)[1:10],
                              object = tan2009r1)
pRolocGUI_SearchResults <- addFeaturesOfInterest(newFeat, pRolocGUI_SearchResults)

Fig. 9: Appearance of search tab (andy2011) in pRolocVis. Search result pRolocVis_Test1 contains one feature of interest (PGAP1_Human)

2.7. data

The tab data is the last tab for pRolocVis and pRolocComp.

Fig. 10: Appearance of data tab in pRolocVis

For pRolocVis the drop-down menu lists all the names of assigned MSnSet objects to the function. Depending if a named list - containing MSnSet instances - is uploaded or not these names will be used or automatic names will be created (object 1 ... object n, where n is the length of the list). In addition, the entry upload is listed at the bottom of the drop-down menu. Selecting upload allows to use a MSnSet available in a .Rda or .Rdata file after uploading it to the application. Clicking on Browse... will open a dialog window with which you can select a file containing a saved MSnSet and load it to pRolocVis. Make sure if you want to use a MSnSet from a .Rda or .Rdata file to select upload in the drop-down menu.

Fig. 11: File upload in pRolocVis

pRolocVis will print a message if there are any conflicts with the uploaded file. If so, either the assigned object itself or the first element in the list will be used instead of the uploaded .Rda or .Rdata file.

When analysing multiple data sets in one pRolocVis session the selected features will be (irreversibly) deleted when changing from one MSnSet to the other! It is therefore highly recommended if this behaviour is not intended to save selected features by using the functionality to save features in a FoICollection first before changing the MSnSet (see 2.6. search for further details).

For pRolocComp the main panel shows a summary matrix of common and unique feature names consisting of common and unique features for all feature names of the two MSnSet instances (row all) and for distinct subsets which are defined by the feature variable names. The feature variable names can be defined by choosing appropriately the drop-down menus marker object 1 and marker object 2 which are located in the section Summary matrix in the sidebar panel (enter ?FeatComp to retrieve more information about the summary matrix). The matrix in the main panel will show only features for distinct subsets when both drop-down menus do not have none as their value, otherwise only the row all will be printed. Given tan2009r1 and tan2009r2 as input for pRolocComp, when selecting markers both for marker object 1 and marker object 2 the matrix will consist of six rows (all, unknown, ER, mitochondrion, Golgi, PM). The table can be interpreted as follows: For the row all we have 545 common features, i.e. there are 545 features in the two MSnSets which have identical feature names. In tan2009r1 however, there are 343 feature names which do not exist in tan2009r1 and 326 feature names are unique for tan2009r2. For markers the matrix tells us that features were assigned to 4 organelles on an experimental evidence (ER, mitochondrion, Golgi, PM) for 833 features in both MSnSets the affiliation to a certain organelle is unknown. In total 20 features are assigned to ER for tan2009r1, therefrom 16 features are also present in tan2009r2, however, 4 are unique to tan2009r1. The interpretation of the other rows can be deduced in the same way.

Features can be selected and displayed in the other tabs, e.g. highlighted on the PCA plot. The selection is conducted in the section Selection. It is made by selecting a marker via the drop-down menu select marker and the radio buttons underneath. The selected row, column and the number of features which is comprised in these categories will be displayed in bold in the summary matrix. By pressing Submit selection the features will be saved internally, thus, available for displaying/use in other tabs. The button Submit selection will be only shown when the features are not already stored internally). If the features are selected there will be the button Undo selection which allows to remove features from the internal selection.

Fig. 12: Appearance of data tab in pRolocComp. The row ER and column unique1 which comprises 4 features are selected

The tab Data in pRolocCompalso allows for subsetting the data sets by choosing accordingly common, unique and common & unique (default) in the section Submit MSnSets. The subset is made on the basis of the feature names of the two MSnSet instances. Selecting common will use only features which occur in both MSnSets, unique will use no common features and common & unique will use all feature names for each MSnSet.


  1. Display selection widget

The Display selection widget is probably the most important implementation in pRolocVis/pRolocComp and allows for identifying features. You can do this by selecting points in the PCA plot, clicking on features in the tab protein profiles, using past searches and/or querying for features in the MSnSet data. In pRolocCompthere is in addition the possibility to retrieve features from the summary matrix in the tab data.

There are four (pRolocVis) or five (pRolocComp) check boxes in the Display selection widget which represent the before mentioned ways of searching features in the MSnSet. To activate the search for one specific method click on the check box left of its description. It is also possible to select more than one at a time which allows for greater flexibility with regard to information retrieval. To irreversibly reset the selection press Clear features (only shown when features are selected).

3.1. PCA

If you decide to identify proteins in the PCA plot, change to the tab PCA and start clicking on features in the PCA plot (tip: the zoom function may be of great expedient). When hovering over the PCA plot the feature meta-data of the nearest feature will be displayed below the plot. The check box will be checked when you start clicking in the PCA plot. As soon as you have clicked on a feature it will be marked with a black circle around it (or a blueish if colour is set to none). If you have selected a feature by accident or want to deselect it, just click again on the feature and it will be deselected.

Selecting features works also in pRolocComp: just click on features in one of the PCA plots will also highlight the same feature in the other plot if it is present. If you want to analyse features which are only common in the two MSnSet instances, go to tab Data and select common in the radio button list Features used.

Fig. 13: Display of selected features in PCA plot (andy2011) for pRolocVis. The features selected originate from selection in the PCA plot

There are two possibilities to deselect all selected features: If you decide to remove all your features click on Clear features (the button will only show up if features were already selected). Please keep in mind that this step once carried out is irreversible and will delete features selected in protein profiles, (summary matrix) and query as well. Besides that you are also able to simply blind out the selected features by deselecting the check box left of PCA in the Display selection widget. Internally, the features are still stored, i.e. by clicking again on the check box you will see the selections again. Clicking on new proteins in the PCA plot will not check the check box again, so you have to do this manually. The features selected are shared between the different tabs. Click on the tabs quantitation and feature meta-data to have a look upon information about the selected features. For the case where you see all features in the data table change the radio buttons settings from all to selected at the lowermost widget in the sidebar. Here again, you can compose the features from different sources (PCA, protein profiles, saved searches and the text-based query search).

If you display protein profiles in the tab protein profiles selected features will be displayed by black lines on all plots drawn.

Fig. 14: Display of selected features in protein profiles plot for pRolocVis. The features selected originate from selection in PCA plot (andy2011)

3.2. protein profiles

In principle the search for features in protein profiles is in accordance with the search in the PCA plot. Though, bear in mind that you are only able to select features when 1 is selected in the drop-down list number of plots to display. Hovering over the plot will display the feature meta-data of the nearest protein below the plot. Clicking on (or near) the points in the plot will select, clicking another time will deselect features. The features will only be shown when the check box left of protein profiles is activated. Note, that you can only select or deselect features whose protein profiles are displayed in a transparent manner on the plot.

For pRolocComp selecting features works in the same way: click on features in one of the protein profile plots (make sure that 1 is selected in the drop-down list number of plots to display), thus highlighting the same feature in the other plot if the same feature name is present. If you want to analyse features which are only common in the two MSnSet instances, go to tab data and select common in the radio button list Features used.

Fig. 15: Display of selected features in protein profiles plot for pRolocVis. The features selected originate from selection in protein profiles plot (andy2011)

3.3. saved searches

Clicking on the check box to the left of saved searches will load the selected features of the class FeaturesOfInterest. These will be displayed in the PCA plot, in the plots for protein profiles (depending on the displayed features) and will be available in the tabs quantitation and feature meta-data for information retrieval. Add FeaturesOfInterest by clicking on the respective features in the tab search in the multiple drop-down list; thus accordingly altering the selected features in the Display selection widget context. Each FeaturesOfInterestinstance will be highlighted in a different colour to distinguish easily between them.

In pRolocVis/pRolocComp there is no functionality implemented to remove features from the object pRolocGUI_SearchResults in the global environment. The authors decided that it is not the task of a GUI to fulfil the requirements of this kind of data manipulation in a GUI, hence, the execution of removing features of interests belongs to the field of the users responsibility.

### 3.4. summary matrix (pRolocComp only)

In pRolocComp another way of selecting and displaying features is possible via the tab data. Features can be selected and internally stored by selecting "all"/a marker via the drop-down menu select marker - this will change the selected row - and one of the radio buttons underneath - this will change the column. By pressing the Submit selection button the features comprised in these categories will be stored internally. The button will be only shown when the features are not already stored internally and can be displayed/used in the other tabs. The selected row, column and the number of features which is comprised in these categories will be displayed in bold in the summary matrix. If features were submitted another button will be present which allows to remove the features from the internal selection, the Undo selection button.

When submitting features from the columns unique1 and unique2, the feature names will only be saved internally to the correspondent MSnSet and displayed there accordingly. This is done because it is possible that the same feature name exists in the other MSnSet but is not assigned to the organelle.

As an example, for the two MSnSetinstances tan2009r1 and tan2009r2 a selection of markers and markers as marker object 1 and marker object 2, when selecting mitochondrion in select marker and the radio button next to common will be equivalent to and will contain the following features
data(tan2009r1, package = "pRolocdata")
data(tan2009r2, package = "pRolocdata")
featcomp <- compfnames(x = tan2009r1, y = tan2009r2, fcol1 = "markers", fcol2 = "markers", verbose = FALSE)
## the fourth element in the list bears information about features which are assigned to 
## "mitochondrion" in "markers", we can access all common features for tan2009r1 and tan2009r2 by
feat <- slot(featcomp[[4]], "common") 
feat
## [1] "Q9V3V6" "Q95083" "Q9V3P6" "P55035" "Q9VW54"

The selection means that for the marker object markers there are nine common features comprised in the two data sets which are assigned to Mitochondrion in the column markers of the feature meta-data. Going to the tab feature meta-data will list these features as well as meta-data.

The stored features will only displayed when the check box next to summary matrix in the Display selection(#display) widget is selected.

3.5. query feature-meta data

The Display selection widget offers the opportunity to query the feature meta-data of the MSnSet for levels. The drop-down list consists of the item protein, which will by definition the feature names and depending on the data accession number, protein ID, protein description, assigned markers (varying on the underlying MSnSet).

For demonstration purposes we will use pRolocVis to select and display features by using the query functionality. Keep in mind, to adjust the selection of the radio buttons next to the appropriate MSnSet when using pRolocComp: Accordingly to the selected MSnSet the list of feature variables is rendered.

Let's assume we want to look at andy2011 which was derived from experiments of Breckels et al. (2013) for all proteins which are assigned by experimental evidence to the organelle plasma membrane. We ensure ourselves that andy2011 is selected in the tab data and change to a tab where the Display selection widget is loaded. We select marker in the upper drop-down list (for we are looking for organelles assigned to marker proteins). In the next drop-down list below we select PM which codes for plasma membrane. Next, we click on Submit selection, which will highlight all features which are assigned to PM for the variable name marker (the button only appears in the application when the corresponding proteins do not exist in the selection). To remove the selected features from the internal assignment we have to either reset the search by clicking on Clear features or click on Undo selection. The latter will only remove the current selection of features, while the former will clear all features (also these of PCA, protein profiles (and data). Of course, we can also add other features: If we want to add all features which are assigned to the Golgi apparatus we simply select Golgi in the lower drop-down list and click on Submit selection to save internally the selected features.

It is relatively easy to find levels when the drop-down list for these levels. But how should we proceed when we want to look for a special protein, e.g. ACADV? The drop-down list for the variable name protein is very long and it is time consuming to scroll through the whole list and look for our protein of interest. Therefore, we can just enter ACADV in the text input field Search for in between the two drop-down lists and we will get the protein of interest (we are also able to query for protein names which have the string AC in their name which will limit the drop-down list to all proteins which have this specific string). By clicking on Submit selection we save internally the selected feature(s).

Fig. 16: Query for proteins in pRolocVis which contain the string "AC". This narrows the features in the drop-down list accordingly (andy2011)


  1. References

[1] L. M. Breckels, L. Gatto, A. Christoforou, A. J. Groen, et al. "The effect of organelle discovery upon sub-cellular protein localisation". In: J Proteomics 88 (2013), pp. 129-140. DOI: 10.1016/j.jprot.2013.02.019. URL: http://www.ncbl.nlm.nlh.gov/pubmed/23523649.

[2] T. P. Dunkley, S. Hester, I. P. Shadforth, J. Runions, et al. "Mapping the arabidopsis organelle proteome". In: Proc Natl Acad Sci USA 103.17 (2006), pp. 6518-6523. DOI: 10.1073/pnas.0506958103. URL: http://dx.doi.org/10.1073/pnas.0506958103..

[3] L. J. Foster, C. L. d. Hoog, Y. Zhang, Y. Zhang, et al. "A mammalian organelle map by protein correlation profiling". In: Cell 125.1 (2006), pp. 187-199. DOI: 10.1016/j.cell.2006.03.022. URL: http://dx.doi.org/10.1016/j.cell.2006.03.022.

[4] L. Gatto and L. M. Breckels. pRolocdata: Data accompanying the pRoloc package. R package version 1.2.0. 2014.

[5] L. Gatto, L. M. Breckels, S. Wieczorek, T. Burger, et al. "Mass-spectrometry based spatial proteomics data analysis using pRoloc and pRolocdata". In: Bioinformatics 30.9 (2014), pp. 1322-1324. DOI: 10.1093/bioinformatics/btu013. URL: http://bioinformatics.oxfordjournals.org/content/30/9/1322.

[6] L. Gatto and K. S. Lilley. "MSnbase - an R/Bioconductor package for isobaric tagged mass spectrometry data visualization, processing and quantitation". In: Bioinformatics 28.2 (2012), pp. 288-289. DOI: 10.1093/bioinformatics/btr645. URL: http://www.ncbi.nlm.nih.govpubmed/22113085.

[7] D. J. Tan, H. Dvinge, A. Christoforou, P. Bertone, et al. "Mapping organelle proteins and protein complexes in Drosophila melanogaster". In: J. Proteome Res. 8.6 (2009), pp. 2667-2678. DOI: 10.1021/pr800866n. URL: http://pubs.acs.org/doi/abs/10.1021/pr800866n.