geWorkbench

Feature requests for geWorkbench

General Features

• Mantis 0000480: Workflow support

Would be nice to have true workflow support for regular users (without the need for scripting). (agreed: Bernd)

• Mantis 0000740: selections in Project folder

When selecting the Workspace, or Project from another object the views that relate to that latter object are still available. It is impossible to know what is being calculated.

When leaving an object the display should be reset to an empty state

• Not in Mantis:

There are problems related to memory management. geWorkbench uses too much memory for storing Microarray data internally. When loading 9 U133 chips one needs more than 512MB.

• Mantis 0000114: Persisting configuration settings designated through the visual builder

At present, changes to the start up configuration effected during the application execution (through the Visual Builder) do not get persisted: when the application is launched the original start up configuration is used again.

It is desirable that changes introduced by the Visual Builder be persisted in the application's configuration file so that they can be "remembered" at the next launch.

• Mantis 0000115: Choosing among available application "flavors"

During application launch, a start up window (which the user can opt to hide for subsequent application invocations) should prompt the user to select one among the available application configuration "flavors". This functionality should be available not only at start up but also from within the application, through a "Preferences" type of menu option.

• Mantis 0000077: Context sensitive online help

We need to make online help available in a context sensitive manner: from within a component, a user should be able (using F1 or by pressing a Help button) get access to the online help associated with the component.

• Mantis 0000156: Extension of event exchange model
  Entering this here as a placeholder regarding the proposed re-engineering of the event model. Some of the design suggestions that have been mentioned include:
  • Making it a rule that event data are always interfaces.
  • Removing the need to specify within the throwEvent() method the listener interface and the method to invoke in that interface.
  • Using annotations (or some other mechanism) in order to provide a direct reference to a service provider so that methods can be invoked directly rather than through event exchanges.

• Mantis 0000157: Extend framework to bring to focus components that receive appropriate events

An issue with the GUI is that one needs to know which components respond to which events in order to inspect the results of some action. E.g., when executing the hierarchical clustering analysis and in order to review the results one needs to explicitly select the Dendrogram tab.

The framework needs to be extended so that components can gain the focus as appropriate when they receive certain events.

• Mantis 0000173: UI Tab Display.

Tabs should be displayed in alphabetical order. It's difficult to find tabs as they are ordered currently.

If the user cannot view all the tabs without scrolling, the tab headers should become drop down values.

• Mantis 0000478: File save/load operations are memoryless in terms of last directory used.
  There are many places in the app where data are being loaded from or saved to disc. When such an operation is used for the *second* time the app should remember the directory the user navigated to the *first* time. Some specific examples where this is not the case:
  • Exporting an image node from within the projects folders.
  • Saving a data node from within the projects folders.
  • Saving a panel from the Markers component.

• Mantis 0000635: columns can be moved (applies to all tables)

when moving a column (which is not neccessarly bad) to the first position (row names) the row name is exchanged with a value but has this "button" look....

• Mantis 0000547: Inaccurate Error Message

When connecting to a MAGE-OM Server the error message is "can not connect to server" even when it does connect and authenicate the user.

I changed the related code to reflect the real error. But right now I have no available caArray RMI server to test this bug, if John can give me access to his local caArray server, I can confirm the bug is fixed or need more work.

---

Programmatic features

• Mantis 0000400: Inconsistent t-test results based on phenotype/panel activation order

There should have been no significant genes found once the marker selection had been made but the volcano plot was erroneously displaying the genes that had been selected as being significant. This is due to an inconsistency in the way the markers() method is implemented in CSMicroarraySetView. For the time being I've routed around this method but it should probably be changed at some point.

---

Menu

• Mantis 0000166: main menubar controls for Project Folders Area are confusing.

The problem is caused by the differences in type and semantics between workspace folder and project folders. For example, The "File" submenu offers "Open" and the pop-up choices are "File" or "Workspace". Select Workspace and you are offered a dialog box. Select "File" and you are told you must first select a project node. Or if you select "New" (an adjective) from the submenu, you get a choice of "Workspace" or "Project". Select "Workspace" and the current workspace folder is cleared of its contents. Thus New Workspace actually REMOVES the Workspace folder, whereas the submenu item "Remove" does not include "workspace" as an option. Select "Project" and a new project folder is added without replacing an existing one. I think the whole thing could be made more obvious and logical by making the File submenu items "Workspace" and "Projects" and perhaps "Files". Then the choices under each could be actions (verbs) - e.g Open, Remove, Load, Delete, Save, etc. In other words, you consider the folder that you want to do something to, select it in the file menu and choose the action that you want to take. Also the other file items "Export" etc. have nothing to do with the Project Folders Area and are positioned awkwardly.

---

File

• Mantis 0000345: File Loading Confusion

No choice to list all files in directory. The file selection box should contain an option to see all the files in a directory, otherwise unclear which file postfix goes with which loading option.

• Mantis 0000347: Prompt to Save

No prompt to save workspace when application exits, there should also be a way to exit the application from the file menu.

• Mantis 0000452: filtered datasets misrecognized on read in

If a dataset is filtered, written out, and then read in again, the "magic marker" that defined the chip type may be deleted, and the dataset can be recognized as another entirely. For example, my HG_U95 dataset, after filtering out half the markers, was recognized as HG_133_Plus2.

• Mantis 0000479: Enable the "Export" functionality for data nodes.

This functionality is supposed to be similar to the "Open file" functionality, where a number of pluggable export filters can be used to export a dataset into another format. At the vry least we should support exporting microarray data into the Cluster (http://rana.lbl.gov/EisenSoftware.htm) [^] format.

• Mantis 0000484: At the File menu, an "Exit" menu item should be provided.

Many many other softwares do have it. Maybe an option for saving the workspace should be provided before exit.

• Mantis 0000465: Support loading Affymetrix .CEL and .CHP formatted files

geWorkbench 1.0/caWorkbench v3.0 are capable of reading in only the .txt version of Affy files. Many users (including Northwestern) have indicated it would be tremendously beneficial to support .CEL, .CHP files as well. The former contain probe level data, rather than the probeset level data which the workbench is currently designed to handle. As such, loading of .CEL data will probably have to be coupled with the immediate execution of a normalizer capable of translating probe level data to probeset level data. => CEL is already working

• Mantis 0000418: Chip type selection dialog appears with dataset switch

If the application doesn't recognize the chip type of a dataset it will ask you to select it, but it will keep asking you every time you switch to the dataset even once after you've picked that type once.

• Mantis 0000549: GUI Poorly Formatted

The GUI for connecting to a remote ca-array server is poorly laid out - Port and Protocol should be on the same line, URL should have its own line, there should be larger room for User Name and Password.

---

---

---

Commands

---

Panels

• Panels should be called sets

---

---

---

---

Help

• Mantis 0000466: Add "shortcuts" screen

Add a "shortcuts" screen (maybe under the "Help" main menu item?) to list available shortcuts like the very useful F12.

• Mantis 0000548: No documentation pointer to MAGE-OM Setup

Although there is documentation in 1.03 geWorkBench describing how to connect to a MAGE-OM server, it is not detailed, does not describe the protocols and has no pointer to the MAGE-OM documentation.

• Mantis 0000511: Help topics not alphabetical

The main help topics in the online help are not in alphabetical order.

---

---

Project Folders

• Mantis 0000739: right click image nodes should show option to export

if there is an image in the project panel there should be an additional option to export the image. Now one has to go through the File menu.

• Mantis 0000633: copy function

Since all the operations are destroying the original data I believe it could be useful to have a copy function that copies a state of a data set to a new data set.

• Mantis 0000643: ask if should really remove

I think it is good practice to ask the user if he really wants to remove an object. Accidents happen and if you put a lot of work into one it is rather disappointing if you accidentally remove your work... (This is only ment for projects and dataset, not for images or such)

• Mantis 0000571: sequence selections under Project folder

1. load two or more sequences in different project files. 2. by holding the ctlr key select more than one sequence. => two sequences are selected, but you cannot do anything the both together. How do you combine different sequence into on project such that you can select from the combined sequences?

---

---

---

---

Arrays/Phenotypes

• Mantis 0000393: Inconsistent GUI element

there is no save/load functionality for arrays as there is for sets

• Mantis 0000415: Panel Label is not prepopulated with last entry.

I performed the following step: - Loaded webmatrix.exp - Added 3 arrays to the selection panel - Selected an array> Add to Panel and typed over the "Section' and entered Panel A - Then another array > Add to Panel. The Panel label text box was no longer prepopulated.

Expected Result: The Panel label should be prepopulated with the last label created (Panel A).

---

---

Sets

• Mantis 0000381: Renaming a label erases its class (case, control, etc.) status.

Renaming a label erases its class (case, control, etc.) status. To reproduce, set a label's class to 'case', then rename that label.

• Mantis 0000451: if filter after clustering, dendrograms become invalid

1. Load a dataset and perform a hierarchical clustering. 2. View in the dendrogram panel. 3. Now filter out some values so that the size of the dataset is decreased. 4. run hierarchical clustering again. 5. View in the dendrogram panel. Now select the first cluster output in the project panel. 6. Because the reduced dataset does not match the orignial cluster, the drawing fails with numerous bad consequences across the interface

How to handle changes in dataset size is a design consideration that must be worked out, for example by 1. duplicating the orignal dataset so that dependent datasets remain valid, or 2. remove the dependent datasets that have become invalid. User should be prompted before such filtering on a dataset that has dependent datasets.

---

---

FASTA sequences

Promoter Panel

• Mantis 0000653: position of promoter panel

To me the promoter panel is more an analysis module and should be in the lower right area (analysis area) rather than in the viewing area (upper right).

• Mantis 0000655: after scan displayed panel is changed

The active/displayed view of the promoter panel is changed to Sequence-Line view whenever a scan is performed.

Especially when changing parameters it can be really annoying.

---

Sequence Panel

• Mantis 0000575: edited sequence causes problems

1. load attached document 2. open sequence view 3. edit sequence (right click on sequence in Project folder, view in editor) 4. remove 7 nucleotide (AATAATT) around position 380 (the line should now be 70 nucleotides long as all the others) 5. activate the sequence (line view) in the sequence panel such that the sequence is shown below the line view 6. click on the arrow bars => sequence disappears.

Also the sequence from the modified filed is not the one show. The original sequence is shown.

---

---

Simulation

• Mantis 0000608: why can a simulation be performed on an image?

Simulation can be used for any object, why? Can it be at least moved to the beginning? So it is not the panel that pops up all the time first?

---

---

---

---

5. Sequence alignment

• Mantis 0000544: fastacmd support doesn't exist

I don't know if there's a plan to implement fastacmd in geworkbench, but if there is, would I find it in sequence alignment? What options will be available? => Don't know why we would support this??

---

BLAST

• related to 0000541: blastall is not complete

using local / own databases

---

---

---

Pattern discovery

• Mantis 0000496: sequence masking needed

A major planned workflow of geWorkbench is to be able to search for patterns in sequences surrounding co-regulated genes. However, no provision is made to mask or deal with masked sequences. Many patterns found in unmasked genomic sequence can be expected to be in repeated elements of various types (or does Splash already have a way of dealing with this?). Solving this may involve enhancements to both the Pattern Discovery module and perhaps to the Splash server. The Splash server currently is believed to recognize mask characters N and #.

Searching for regulatory patterns is potentially very sensitive to the details of how it is carried out, as such patterns may be short and embedded in regions of repeated sequence. The greatest possible degree of flexibility and control may be required.

Suggestions: We will assume we can obtain from the Sequence Retriever sequence that is either unmasked or is reversibly masked, that is with small letters denoting masked seqeunce and capitals denoting unmasked.

The following options cold be considered: 1. If lower-case-masked sequence is available, provide the option to convert masked characters to Ns before sending to the Splash server.

2. We could host a repeatmasker server (www.repeatmasker.org). We could then mask sequence as desired, e.g. mask low complexity but leave in LINE elements. There are many considerations and subtleties discussed at repeatmasker.org.

3. The Splash server itself could be enhanced to offer simple masking, along the same lines as the BLAST programs do.

4. We should give some indication of the kinds of statistics one could expect from true conserved patterns vs those arising from repeats. Can they be distinguished?

• Mantis 0000665: conceptional inconsistency

Pattern discovery is the only component I came across so far that has “Pattern discovery” (itself) in the analysis section. All other modules just have “Experiment Info”, etc. and in the visualization area the results. Actually I do like the idea of having “itself” still in the analysis section because this way the parameters are still stored. But it is still inconsistent.

---

---

---

---

---

---

---

Microarray data

The following panels/functionalities are requested specifically for Microarray data:

---

---

---

---

Tabular Microarray Viewer

• Mantis 0000627: selecting arrays and markers

It would be nice if markers and arrays could be selected/highlighted by double clicking on row / column headers

---

---

---

Microarray viewer

• related to (closed) Mantis 0000225: Microarray panel details

It would be nice to have the values (min, max) displayed next to the color legend

• Mantis 0000326: absolute visualization needs viewing controls

If a set of unnormalized Affy data is read in, for example the cardiogenomics dataset, with the tools->preferences->visualization mode set to absolute, the dynamic range of the data is so large that only a few data points appear. Some options to deal with this would be: 1. a control labeled "Display as log transform" 2. an intensity control such as in the color mosaic panel.

• Mantis 0000431: Table View does not sort properly

Sorting by most columns not work properly (does not order numerically or alphabetically) and attempting to sort by pValue appears to do nothing. => also true for any table

• Mantis 0000550: color scale disappears when window small

If the window size is made small enough, the expression color scale disappears, leaving behind just its label. This may be common if people have small screens.

---

Gene Ontology

• Mantis 0000623: Organization of panel is misleading

The Gene Ontology Panel is divided into three sub-panels: TreeView, TableView, P-value Trend. TreeView holds vital information for TableView and P-value: Chip set used, Reference List, selection to choose which of the above, and which class of GO terms is used. Those options/parameters should be visible in all of the three main panels since they heavily rely on them and should be moved outside of the panel structure within the GO panel!

---

---

Reverser Engineering

• Mantis 0000678: Cytoscape - right click within graphics area

right click with the graphics area give you a lot of options including to search Google. Why not search NCBI, Pubmed?

• Mantis 0000679: Cytoscape - resizing nodes

nodes can be resized within cytoscape and the size/shape is retained even when changing the model. There are two ways of getting the original shape back: 1. changing it manually, or 2. by switching to a different Adjacency Matrix and back. This is a bit awkward and inconsistent.

• Mantis 0000677: conceptual inconsistencies

Reverse engineering is located in the visualization area and not in the analysis area. To conceptionally consistent the visualization and parameter selection should be separated.

• Mantis 0000675: Cytoscape - Project Folders

When running multiple analysis creating multiple networks new item in the project folder are created. At the same time new entries in Cytoscape are created. This is redundant and confusing. There can (should) only be one, and it should probably be the Project folder one with all its drawbacks, just because it is more consistent. But do we then need the left panel showing all the networks?

• Mantis 0000674: Cytoscape

Cytoscape comes with a whole set of menues and functionality that is not visible at first, only after resizing the window they come up. => this should be changed

By default cytoscape comes with two panels that can be selected (You will see them after resizing.) They cannot be switched by clicking on the panel heads but only by clicking on the panel names. (The other one was called sim1...)

---

---

---

---

Color Mosaic

• Mantis 0000477: Missing right click actions

CM should have the right click actions consistent with Expression Profiles, EVD,Scatter Plot which supports zoom in/out,save as & print.

=> least have a mouse-over function that displays the chip and value for the spot

• Mantis 0000611: color slider function not obvious (probably not only in color mosaic)

The real values displayed in a heat plot have to be mapped into a color space that is usually in the range between 0 to 1. A threshold has to be defined beyond which all values are mapped to 1 in the color space. By moving the threshold lower differences in the lower magnitudes can be visualized. This does not seem to happen in geWorkbench. I would expect to everything lighting up when I move the slider to one of the extremes. This does not happen.

Since I have seen the behaviour also in the other heat map representation and this is not the obvious behaviour it either has to be clearly documented what is happening or changed to "normal" behaviour

---

Expression Profiles

• Mantis 0000524: no mouse over on expression graph (assigned to John)

Some of the components support displaying the marker name when a data point is moused-over. The Expression Profiles component does not, instead you must click on a line and view its name in the Markers panel. Mouse over would be a good enhancement.

---

---

---

---

---

---

---

---

---

---

---

---

---

---

---

---

---

---

---

---

Analysis

The following analysis functions are requested:

• ANOVA

• mantis 0000307: Analysis should write to dataset history

When any type of analysis is done, basic facts should be written to the Dataset History log, such as whether all markers were used or an activated panel; the number of markers/arrays in the activated panel, and the array type, and a timestamp.

---

Fast Hierarchical clustering

• Mantis 0000148: Euclidean distance shoud use normalized vectors

Data should be transiently normalized

---

---

---

---

SOM

• Mantis 0000685: image snapshot only captures one cluster

It would be nice to capture as well all clusters at once.

---

T-test

• Mantis 0000516: Volcano Plot doesn't know data already log2 normalized

When a t-test has been done in the Analysis component, the results are displayed in the Volcano Plot (note that this component does not have an entry in Mantis yet). The display is on a log2 scale on the X-axis. However, the component does not know if the data has already been log2 transformed. I think it is essentially doing a second log2 transform of the data - that is, instead of just subtracting one log2 value from the other, it is taking the difference and then the log2 of this. Or it is taking a log2 transform of all the data again and then subtracting. Either way, it is displaying a log2(log2 X) value.

The only way to deal with this is to know that the data has been log2 transformed. It would appear that this fact would have to be remembered if done in the normalizer component, or the user should be able to check a box on read-in that indicates the data is log transformed.

We can probably autodetect the log transformed state just on the range of values present in the input file. => auto detection is good, but we need to be able to set it manually, too. (BJ)

---

---

---

---

---

---

---

---

---

---

---

---

Dataset History

• Mantis: 0000300: need array info

We should be able to display information about a microarray - the chip type, the number of markers in the dataset vs in the chip definition.

• Mantis 0000476: Data set annotation missing

Issue: Users analyze data, use activated sets, limit arrays included and none of these steps are captured in the annotation. The system should capture these steps for the user in a log so the researcher is clear on what the results include and how to reproduce.

---