Using ProtPlot for Data Mining Virtual Protein Expression Patterns

TMAP home | Introduction | Using ProtPlot | Menus | Estimating expression | Download |
PRP file format | Revision history | PDF documents | History of ProtPlot | Latest version

Installing ProtPlot

First you need to download and install ProtPlot on your computer. The detailed steps are shown in the PDF Introduction to data-mining with ProtPlot: Screen Shots. This downloads the ProtPlot Java program and the CGAP derived data set of pseudo 2D-gels. If you download the version that includes the Java Virtual Machine (JVM), it will not interact with any other JVM you have installed.

Using ProtPlot

You start ProtPlot by clicking on the "ProtPlot Startup" icon if your computer supports that (Windows, MacOS-X, etc.) or type ProtPlot on the command line for Unix, Linux and other systems.

Once the ProtPlot program is started, it loads the set of PRP files that you downloaded with the ProtPlot program. The virtual protein data for each tissue is used to construct a Master Protein Index where proteins will be present for some tissues and not for others. The data is presented in a pseudo 2D-gel image with the estimated isoelectric point (pI) on the horizontal axis and the molecular mass (Mw) on the vertical axis. Sliders on each of the axes allow you control the minimum and maximum values of pI and Mw displayed and thus the Mw vs. pI scatterplot zoom region you want to select. By clicking on a spot in in the scatterplot, you will display information on that protein. You also define that protein as the current protein. The current protein is used in some of the clustering methods, protein specific reports (Expression Profile report), and the Expression Profile plot. If you have enabled the popup Genomic-ID Web browser and you are connected to the Internet, it will popup a Web page from the selected Genomic database for that protein.

You select various options from the pull-down menus. Some of the more commonly used options are replicated as check-boxes at the bottom of the window.

The Scatterplot Display Mode

There are two primary types of pseudo 2D-gel (Mw vs pI) scatterplot display modes of this derived protein expression data: expression mode or ratio mode. The expression data may be for a single sample (the current sample) or the mean expression of a list of samples (called the expression profile or EP). The ratio data is compute as the ratio of two individual samples called X and Y. Ratio data may alternatively be computed from sets of X samples and sets of Y samples. Generally, one would group a set of samples with similar characteristis together having the same condition (e.g., cancer, normal, etc.). The ratio of X and Y may be single samples in which case the ratio is computed as:
   ratio = (expression X / expression Y)
where expression X (expression Y) is the expression of corresponding proteins. Alternatively, you may compute the ratio of the mean expression of two different sets of samples (the X set and the Y set). The X and Y sets may be thought of as experimental conditions and the members of the sets being "replicates" in some sense. E.g., the X set could be cancer samples and the Y set could be normal samples. The ratio of the X/Y sets for each corresponding protein is computed as
   ratio = (mean X-set expression / mean Y-set expression)

The following shows one of the (Mw vs. pI) scatterplots when the display mode was set to (X-set/Y-set) ratio mode:

It is also possible to create an (X vs Y) scatter plot or (Mean X-set vs. Mean Y-set) scatterplot when the corresponding ratio display mode is set. The following window shows the (Mean X-set vs. Mean Y-set) scatterplot:

The following table summaries the four types of display modes:

Display Mode Current sample Single X/Y X-set/Y-set EP-set
Expression yes no no no
Single samples ratio no yes no no
X-set and Y-set samples ratio no no yes no
Mean Expression no no no yes

Effect of display mode on filtering, clustering and reporting

You select the particular display mode using the Plot menu comands. When you select a particular display mode, it will enable and disable Filter, View, Cluster and Report options depending on the mode. For example, you may only use the t-Test or missing X Y set test if you are in XY-sets ratio mode. You may only perform clustering if you are in EP-set mode. You may change the display mode using the (Plot menu | Show display mode) commands. Alternatively, since it is used so often, there is a checkbox at the bottom of the main window " Use XY-sets" that will toggle between the XY-sets ratio mode and whatever the previous mode you had set.

Selecting Samples

You select samples for the current sample, X sample, Y sample, X-set samples, Y-set samples, and EP-set samples using a popup checkbox list chooser of all samples.

This may be invoked either from the File menu or the pull-down sample selector at the lower-left corner of the main window.

For example, you invoke this chooser for a the specific tissue sample you want to view by using the (File menu | Select samples | Select Current PRP sample). For X (Y) data, you invoke the choosers using (File menu | Select samples | Select X (Y) PRP sample(s)). You may switch between single (X/Y) and (X set/Y set) mode using the (File menu | Select samples | Use Sample X and Y sets else single X and Y samples [CB]) command.

There is an alternative display called the 'Expression Profile' (EP) plot which display a list of a subset of PRP samples for the currently selected protein. You may also display the scatterplot on the mean EP data for all proteins. The EP samples are specified using the (File menu | Select samples | Select Expression List of samples) command.

Listing a report on sample assignments

You may popup a report of the current sample assignments for the: current sample single X sample, single Y sample, X sample set, Y sample set, and EP sample set using the (File menu | Select samples | List sample assignments) command.

Assigning the X-set and Y-set condition names

The default experimental condition names for the X and Y sample sets are 'X set' and 'Y set'. You may change these by the (File menu | Select samples | Assign X (Y) set name) commands.

Status Reporting Window

There is a status popup window that first appears when the program is started and reports the progress while the data is loading. After the data is loaded, it will disappear. You may bring it back at any time by toggling the " Status popup" checkbox at the bottom of the window. You may also press the "Hide" button on the status popup window to make it disappear.

Data Filtering

The pseudo-protein data is passed through a data filter consisting of the intersection of several tests including: pI range, MW range, sample expression range, expression ratio(X/Y) range (either inside or outside the range), t-Test comparing the X and Y sample sets, Kolmogorov-Smirnov test comparing the X and Y sample sets, missing proteins test for X and Y sample sets, tissue type filter, protein family filter [Future], and clustering. The filtering options are selected in the Filter menu. If you are looking at the scatterplot in ratio mode, then you may filter by ratio of X/Y either inside or outside of the ratio range. The missing protein test defines missing as totally missing and present as having at least 'N' samples present. Note that the t-Test and the missing protein test are mutually exclusive in what they are looking for, so using both results in no proteins found.

Saving filtered proteins in sets for use in subsequent data filtering

You may save the set of proteins created by the current data filter settings by pressing the "Save Filter Results" button in the lower-right of the main window. This set of proteins is available for use in future data filtering using the (Filter menu | Filter by AND of Saved Filter proteins [CB]). When you save the state of the ProtPlot database (Filter menu | State | Save State), it will also write out the save protein sets (saved filtered proteins and saved clustered proteins) in the database "Set" folder with ".set" file name extensions.

In the (Filter menu | State | Protein Sets) submenu there are a number of commands to manipulate protein set files. You may individually save (or restore) any particular saved filtered set to (or from) a set file in the "Set" folder. There are also commands to compute the set intersection, union or difference between two protein set files and leave the resulting protein set in the saved Filter set.

Filter dependence on the display mode

Note that the particular filter options available at any time depend on what the current display mode is. The following table shows which options are available for which display modes.

indicate that the command

Filter Name Current sample Single X/Y X-set/Y-set EP-set
> 200K Daltons yes yes yes yes
Tissue type yes yes yes yes
Expression (Ratio) range expression ratio ratio expression
X/Y (inside/outside) range no yes yes no
(X-set, Y-set) t-Test no yes yes no
(X-set, Y-set) KS-Test no yes yes no
(X-set, Y-set) Missing data no yes yes no
At Most (Least) N samples no no yes yes
AND of saved cluster set yes yes yes yes
AND of saved filter set yes yes yes yes

The data-mining 'State'

The current data-mining settings of ProtPlot is called the 'state'. It may be saved in a named startup file called the 'startup state file' in the "State" folder. The "State" folder and other folders used by ProtPlot are found in the directory where you installed ProtPlot. Initially there is no startup state file. If you save the state it creates this file. You may create as many of these saved state files as you want. You may change the file and thus save various combinations of settings of samples for the current, X, Y and expression list of samples. The state also includes the the various filter, view and plot options as well as the pI, Mw, expression, ratio, cluster distance threshold, number samples threshold, p-Value threshold sliders, as well as other settings. The saved Filter and Cluster sets of proteins are also written out as .set files in the "Set" folder when you save the state.

Starting ProtPlot by clicking on the ProtPlot startup icon will not read the state file when it starts up. However, if you have saved a state, clicking on the state file or a shortcut to the state file will cause it to be read when ProtPlot starts up.

You may save the current state using either the (File | State | Save State) command to save it under the current name, or using either the (File | State | Save As State) command to save it under a new name you may specify. Then you may also change the current state using (File | State | Open Statefile) command.

The Molecular Mass vs pI Scatterplot : expression or ratio

There are to types of scatterplots: expression for a single sample or the ratio of 2 samples X and Y. The Plot menu lets you switch the display mode. Ratio mode itself has two types of displays: red(X) + green(Y), or a ratio scale ranging between <1/10 (green) and >10 (red). You may view a popup report of the expression or ratio values for the current protein. If 'Mouse-over' is enabled, then moving the mouse over a spot will show the name of the protein and its associated data. If mouse over is not enabled, then clicking on the spot will show its associated data.

You may scroll the scatterplot in both the pI and Mw axes by adjusting the end-point scrollbars on the corresponding axes. You may display the scatterplot with a log transform of MW by toggling the log MW switch.

The popup plots and scatterplot may be saved as .gif image files which are put into the project's "Report" folder. Similarly, reports are saved as tab-delimited .txt text files in the "Report" folder. Because it prompts you for a file name, you may browse your file system and save the file in another disk location.

X sample(s) vs Y samples scatterplot

If you are in X/Y ratio mode (single X/Y samples or X-set/Y-set samples), you may view a scatterplot of the X vs Y expression data. Enable the XY scatterplot using the (Plot menu | Display (X vs Y) else (Mw vs pI) scatterplot - if ratio mode [CB]). You may zoom the scatterplot just as you do for the (Mw vs pI) scatterplot. The proteins displayed are those passing the data filter that have both X and Y data (i.e., expression is > 0.0).

Expression Profile plot of a specific protein

An expression profile (EP) shows the expression for a particular protein for all samples that have that protein. The (Plot menu | Enable expression profile plot) pops up a EP plot window and displays the EP plot for any protein you select by clicking on it. The relative expression is on the vertical axis and the sample number on the horizontal axis. Pressing on the "Show samples" button pops up a list showing the samples and their order in the plot. Pressing on the "nX" button will toggle through a range of magnifications from 1X through 50X that may be useful in visualizing low values of expression. Clicking on a new spot in the (Mw vs. pI) scatterplot will change the protein being displayed in the EP plot. Within the EP plot display, you may display the sample and expression value for a plotted bar by clicking on the bar (which changes to green with the value in red at the top). You may save the EP plot as a GIF file. You may also click on the display to find out the value and sample. Note: since clustering uses the expression profile, you must be in 'mean EP-set display' mode.

Clustering of expression profiles

You may cluster proteins by the similarity of their expression profiles. First set the plot display mode to "Show mean EP-set samples expression data". The clustering method is selected from the Cluster menu. Currently there is one cluster method. Others are planned.

The cluster distance metric is the 'distance' between two proteins based on their expression profile. The metric may be selected in the Cluster Menu. Currently, there is one clustering method: cluster proteins most similar to the current protein (specified by clicking on a spot in the scatterplot or using the Find Protein by name in the Files menu). It requires you to specify a) the current protein, and b) the threshold distance cutoff. The threshold distance is specified interactively by the "Distance Threshold T" slider. The 'Similar Proteins Cluster' Report will be updated if you change either the current protein or the cluster distance.

The cluster distance metric must be computed in a way to take missing data into account since a simple Eucledian distance can not be used with the type of sparse data present in the ProtPlot database. ProtPlot has several ways to compute the distance metric using various models for handling missing data.

You may save the set of proteins created by the current clustering settings by pressing the "Save Cluster Results" button in the lower-right of the cluster report window. This set of proteins is available for use in future data filtering using the (Filter menu | Filter by AND of Saved Clustered proteins [CB]). When you save the state of the ProtPlot database (Filter menu | State | Save State), it will also save the set of saved clustered proteins in the database "Set" folder. You may restore any particular saved clustered set file.

You may bring up the EP plot window by clicking on the "EP Plot" button and then click on any spot in the scatterplot to see its expression profile. Clicking on the "Scroll Cluster EP Plots" button brings up a scrollable list of expression profiles for just the clustered proteins sorted by similarity.

The following window illustrates the scrollable list of EP plots sorted by the current cluster report similarity.

You may mark the proteins belonging to the cluster in the scatterplot with black boxes by selecting the " View cluster boxes" checkbox at the lower left of the cluster reportwindow. This is illustrated in the following window:

Reports

Various popup report summaries are availble depending on the display mode. All reports are tab-delimited and so may be cut & pasted into MS Excel or other analysis software. Reports also have a 'Save As' button so you can save the data into a tab-delimited file. The default /Report directory is in the directory where you installed ProtPlot. However, you may save it anywhere on your file system. The contents of some reports depends on the particular display mode. This is summarized in the table below.

Filter Name Current sample Single X/Y X-set/Y-set EP-set
Statistics or proteins passing filter SP-ACC/ID, pI, Mw, expression SP-ACC/ID, pI, Mw, X/Y, X, Y expr, Tissues SP-ACC/ID, pI, Mw, mnX/mnY, (mn,sd,cv,n) expr for X- & Y-sets, Tissues. If using t-test then (dF, t-stat, F-stat). If using KS-test then (dF, D-stat) SP-ACC/ID, pI, Mw, (mn,sd,cv,n) exprfor EP-set, Tissues
Expression profiles of proteins passing filter SP-ACC/ID, expr data EP-set SP-ACC/ID, expr data EP-set SP-ACC/ID, expr data EP-set SP-ACC/ID, expr data EP-set
X &Y sets of missing proteins pasing filter no no SP-ACC/ID, (mn,sd,cv,n)for X- & Y-sets no
EP set statistics of proteins passing filter no no no SP-ACC/ID, (mn,sd,cv,n) for EP-set
List of samples in current EP profile {Nbr, sample-name, expression) {Nbr, sample-name, expression) {Nbr, sample-name, expression) {Nbr, sample-name, expression)
List of all sample assignments Current, X, Y, X-set, Y-set, EP-set Current, X, Y, X-set, Y-set, EP-set Current, X, Y, X-set, Y-set, EP-set Current, X, Y, X-set, Y-set, EP-set
List of # proteins/sample {Sample-name, # proteins in sample} {Sample-name, # proteins in sample} {Sample-name, # proteins in sample} {Sample-name, # proteins in sample}
ProtPlot state State State State State

Genomic Databases

If you are connected to the Internet and have enabled ProtPlot to 'Access Web-DB', then clicking on a protein will popup a genomic database entry for that protein. The particular genomic database to use is selected in the Genomic-DB menu.


Djamel Medjahed, LMT, SAIC-Frederick
Peter Lemkin, LECB, NCI-Frederick

Revised: 08-26-2004