iRVis: inteRactive Visualisation of fitness comparisons

You can print this document to have a physical copy to hand while using iRVis.

Introduction

By comparing the fitnesses of independent microbial strains grown under two different environmental or genetic conditions (e.g. query and control conditions) we can rank and compare the strength of gene-gene or gene-environment interactions. Several technologies for measuring microbial fitness exist, including Quantitative Fitness Analysis (QFA), Synthetic Genetic Array (SGA) and liquid culture growth assays. Genome-wide fitness comparisons can become difficult to visualise by static, 2D scatterplot due to the sheer number of strains examined because simultaneous labelling of ~4,000 genes on a single plot is not practical. For example, in Figure 1, the labels for many genes are obscured.

Figure 1: A static fitness plot Fitness plot showing evidence for gene-gene interactions between a query mutation and each of the deletions in the yeast knock-out collection. Mean fitnesses were calculated for each deletion combined with a temperature sensitive query mutation (yku70Δ) and plotted on the y-axis. Mean fitnesses were calculated for each deletion combined with a neutral control mutation (ura3Δ) and plotted on the x-axis. The screens were approximately temperature matched (~37°C). Red and green points indicate genotypes which significantly suppress and enhance the fitness defect of the query mutation, respectively. Blue horizontal and vertical lines intersect at the point corresponding to his3Δ (a wild-type surrogate). Solid grey line is predicted double-mutant fitness, given single deletion fitness and assuming a multiplicative model of genetic independence. Dashed grey line is the line of equal fitness.. This static fitness plot was generated using an early version of the QFA R package and is a reproduction of Figure 2 from Addinall et al., 2011

iRVis is a tool for generating interactive versions of fitness plots (as shown in Figure 1). iRVis allows us to make rapid, visual comparisons between different pairs of QFA experiments and to query plots in real time, aiding analysis and interpretation of the underlying data. iRVis can be applied to any paired sets of control and query fitnesses, including SGA data, or fitnesses derived from liquid growth curves.

We are also currently developing DIXY, a web-based alternative with broadly similar functionality. DIXY has the advantage that it requires no installation, it simply runs in your browser. However, transitions between plots currently run much slower using the web-based tool and so, for the moment, we recommend the R-based version for browsing multiple datasets.

Figure 2: Static versions of iRVis fitness plots Fitnesses from three telomere-related screens plotted against matched controls. Data from Addinall et al., 2011 re-plotted using iRVis. Four functionally related genes are highlighted in purple on each plot and gene names printed. Genes highlighted in green significantly enhance the background mutation phenotype indicated in the y-axis label. Similarly, genes highlighted in red suppress the phenotype. Blue horizontal and vertical lines intersect at the point corresponding to his3Δ (a wild-type surrogate). Solid grey line is predicted double-mutant fitness, given single deletion fitness and assuming a multiplicative model of genetic independence. Dashed grey line is the line of equal fitness. Click on panels to zoom.

On each iRVis plot, the main title indicates the statistic used to summarise replicate fitness observations (e.g. mean or median) and the statistical test used to identify enhancers and suppressors (e.g. Student's t-test or the Wilcoxon test). Similarly, some information about the genotype, environmental conditions, screen IDs and the date the screen was carried out can be found in the axis labels. The interactive features of iRVis can be used to display more experimental metadata in real-time.

Datasets underlying fitness plots

iRVis plots are generated from tab-delimited text files summarising fitnesses and genetic interaction strengths that are output by the report.epi function in the qfa package. Here is an example data file from a comparison between a cdc13-1 screen with a ura3Δ screen carried out at 27°C by QFA, as presented by Addinall et al., 2011 (see Figure 2): DAL_cdc13-1_vs_ura3_27_SDM_rhlk_CTGH_GIS.txt. iRVis generates visual representations of this tabular data, allowing us to highlight genes of interest and to rapidly observe how the fitnesses (and genetic interaction strengths) of highlighted genes change in different backgrounds.

Starting iRVis

First start the R application. Once R is running, type the following command at the command prompt (pressing enter after) to load the qfa package which contains iRVis:

library(qfa)

To view this documentation, we can examine the iRVis vignette, which is a tutorial document. The vignettes available in the qfa package can be listed by executing the following command (please note that R commands are case-sensitive):

browseVignettes("qfa")

Simply click on the iRVis HTML link to access the vignette.

Now that the qfa package is loaded, there are several ways to use the visualisation tool, which will be described later. The qfa package comes with a set of publically available datasets (from Addinall et. al 2011) pre-loaded. To quickly start interacting with these demo datasets, the simplest thing to do is to type the following at the command prompt (followed by enter):

iRVisDemo()

Each time iRVis starts, it displays some useful summaries and file locations on the console. This information includes the following:

Managing the console and the plotting window

It is important to note that iRVis consists of two windows: the text-based console and a graphics window for displaying plots. Upon starting iRVis, a plotting window should appear alongside the R text console and a brief summary of instructions for using the keyboard and mouse to interact with iRVis should be listed in the console window. The user can interact with both of these windows and may need to switch back and forth between them while using iRVis. Note that you may need to click on the console window to put it into focus before you can interact with it (e.g. when typing in Gene names etc.). Similarly, you may have to click on the plotting window to put it into focus before attempting to use any keyboard plotting commands. More detailed instructions below.

Interacting with plots

Users can interact with plots in several ways, using the keyboard and mouse. The full set of controls are described separately below, but here is an overview of some of the functionality of iRVis. Users can:

There are many different approaches you can take to view your fitness data with iRVis. We suggest you explore the functionality available and discover what is most useful for you.

Mouse interaction

Keyboard interaction

Browsing fitness plots

Highlighting groups of genes

Highlighting significant interactions

Switching between plot styles

Saving plots to file

Experimental metadata & selected genes

Quitting iRVis

Different views of the data (press 'l')

Pressing 'l' toggles between the standard view of QFA data: the fitness scatterplot and alternative rank plot views. In rank plots we plot various measures of interaction or fitness against some ranking of the genes on the x-axis. There are several different options available for sorting or ranking the points.

By repeatedly pressing 'i' in rank mode, the user can toggle between ranking genes as follows:

By repeatedly pressing 'k' in rank mode, the user can toggle between y-axis variables as follows:

The current ranking is listed in the x-axis label of rank plots.

Figure 3: Static versions of nine different iRVis views of a single dataset Fitness plot and eight (of a possible 20) rank plots of the cdc13-1 vs. ura3Δ experiment presented in panel A of Figure 2.

Highlighting other groups of genes

Three default lists of functionally related gene complexes are presented within the qfa package:

  1. buildComplexes A list of about 500 related groups of S. cerevisiae genes including telomere-related genes that are of interest to the Lydall lab together with hits from telomere-length screens by Askree et al. (2004) & Meng et al. (2009), replicative senescence screens by Chang et al.(2010), a chromosome instability screen by Stirling et al. (2011) and a manually curated list of functionally related complexes by Benschop et al. (2010)
  2. buildGO A list of all S. cerevisiae genes that are co-annotated with each of about 4,000 GO terms from SGD.
  3. buildPombeGO A list of all S. pombe genes that are co-annotated with each of about 400 GO terms from the pombase database.

Executing iRVisDemo(buildComplexes) allows highlighting of groups of genes from list 1 above. We imagine that this list will be the most useful for users and so we have allowed the following shorthand for starting iRVis with this list loaded (please note that R commands are case-sensitive):

iRVisDemo()

If you would like to use a version of iRVis with genes co-annotated by specific GO terms highlighted instead (list 2), first quit iRVis (by pressing 'q') and type the following command in the text console (followed by enter):

iRVisDemo(buildGO)

Similarly, if you would like to highlight genes co-annotated with GO terms in a S. pombe dataset, you can use the following command:

iRVisDemo(buildPombeGO)

Note that there are about 4,000 GO terms to browse through in S. cerevisiae datasets…

Loading other datasets

You might receive different datasets in the form of R packages from the Newcastle University High-Throughput Service or from the Lydall lab. We find that grouping QFA experimental results together in separate, data-only R packages is a useful way to distribute datasets separately and privately. In order to browse the QFA data within a data-only R package, you will need to install the data package (in addition to the qfa package you have already installed). For example, you may receive a package file named qfaDAL_0.0-9.zip (Windows) or qfaDAL_0.0-9.tar.gz (OSX and Linux). Some installation instructions for Microsoft Windows and OSX users can be found here.

Once the data package is installed, you will need to load it. By loading the data package, R will automatically load the qfa package (and therefore iRVis). In order to load the data package, note its name: the part of the package filename before the underscore: qfaDAL in the example above. To load a package named qfaDAL, start the R application and type the following into the terminal, followed by enter (please note that R commands are case-sensitive):

library(qfaDAL)

Once the data package is loaded you can launch iRVis as follows:

visAll()

As we did for iRVisDemo(), we can load an alternative list of groups of genes co-annotated with one of 4,000 different GO terms:

visAll(buildGO)

The demo datasets from Addinall et al. 2011, included in the qfa package, will always be present for comparison.

You can print this document to have a physical copy to hand while using iRVis.