maxdView is a system for analysis and visualisation of expression data. The design is focused on providing a comprehensive model of expression data and a flexible plug-in architecture to allow functionality to be added as desired, and as new techniques are developed.

• Data Model

  • An expression data class which represent results from one or more hybridisations and any associated clusters of genes, probes and spots.
  • Flexible naming scheme allowing unlimited tags to be linked to names
  • An extensive API for manipulating expression, cluster and annotation data programatically.
  • Textual Annotation can be gathered from multiple external sources, cached locally and accessed via built-in browser.
  • Support for derived data such as probability and error values
  • Remote access to data through Java's RMI protocol
  • Support for application wrapping so that maxdView can be embedded in another Java application
    (see NCGR's ISYS)
  • Extensive help documentation accesed via built in context sensitive browser (includes search facility).

• Import/Export

  • Open file format using XML syntax
  • Database load for retrieving data from any maxdLoad2 database
  • Data import facility for loading data from ASCII text files
  • Data export as ASCII text with a variety of formatting

• Transformation

  • Data normalisation using a variety of standard and novel methods
  • Sophisticated mathematical manipulations using a convenient calculator style interface
  • Sorting of data based on numerical values, textual annotation or cluster occupancy
  • 'Principal Components Analaysis' for data simplication
  • Transparent interface to external programs or scripts
  • Grouping and clustering of data based on expression profiles
  • Interfaces to the XCluster and Weka machine learning packages.
  • Dynamic compilation and execution of arbitrary Java code-fragments

• Filtering

  • Filtering based on complex expressions relating one or more combinations of Measurements
  • Filtering using regular expressions to search names and annotation
  • Filtering by expression profile similarity
  • Filtering based on cluster occupancy

• Visualisation

  • The standard table layout (with optional cluster trees)
  • Data quality analysis plots including Zipf's and Benford's and a novel statistical quality control algorithm
  • 2D correlation plot with overlay of cluster data
  • Expression profile viewer, optionally organised by cluster hierarchy
  • Multidimensional plots (with many projection options)
  • Distribution histograms for one or more data sets
  • Interactive 3D correlation plot with configurable colour and size mappings

As the system has been implemented in Java, dynamic class loading can be used to make the system highly modular; components of the system, such as a data loading component, can be added whilst the application is running. This is particularly useful during development, as new modules can be tested, the code changed, and recompiled and then the module can be reloaded into the system. Dynamic loading is also used to allow the user to directly enter code into a window and have it be executed within the system. By using the data API in this way, users can perform arbitrary numerical processing of the data without too much effort.

The current version cna be downloaded from: http://bioinf.man.ac.uk/microarray/maxd/download.html.