The goals of Gemma are to:

• Aggregate public gene expression studies for the purposes of reanalysis. Our current focus is on human, mouse and rat nervous system data.
  
• Perform analyses on the expression data and store the results for rapid and easy access.
  
• Reanalyze probe sequences to allow consistent cross-platform comparisons.
  
• To serve as a framework for the development and evaluation of novel algorithms for expression analysis and meta-analysis.
  
• Support a range of analytical approaches including but not limited to coexpression and differential expression.
  
• Support a variety of expression technologies, including Affymetrix, Illumina and other oligonucleotide arrays, one channel and ratiometric cDNA arrays, and (planned) SAGE or other sequence-based techniques.
  
• Provide consistent annotation/descriptions of data sets to enhance usability of the data.
  
• Provide external sites with processed data, web services, software and other means of including Gemma tools and data in other systems.
  
• Allow users to upload expression data for inclusion in Gemma, either privately for their own analyses or to be made public. (This is a feature under development!)
  
• Provide a software system that can be installed and used by others.
  

To these ends, Gemma consists of:

• A database and software framework for storing and accessing expression data with a focus on analysis. The software is released under the Apache 2.0 license, which is liberal enough to allow commercialization, modification, and reuse by anybody.
  
• A set of administrative and curation tools for loading, tracking, processing, annotating, analysis and quality control of expression data.
  
• A sequence analysis pipeline for probe sequences.
  
• Implementations of algorithms for analysis of expression data, in some cases using integration with "R";
  

The main functionality of Gemma
For users of the web site, Gemma has two main functions, which are supported by various searching and browsing functionality.

• Coexpression meta-analysis: The entry point for coexpression analysis is here. The methods used for this "recurring coexpression" analysis are essentially as described in Lee et al, 2004. Briefly, each data set is preprocessed and analyzed to identify pairs of genes that are strongly coexpressed. The on-line analysis searches these results for pairs of genes that are coexpressed in multiple data sets, starting from a gene you choose.
  
• Differential expression analysis access:We are performing differential expression analysis on most of the data sets in Gemma. Currently these results are available from a search form or integrated into the results.