GAIN is based on interaction information between three attributes; in
this case, between two single nucleotide polymporphisms (SNPs) and a
class or phenotype attribute. Interaction information is the gain in
phenotype information obtained by considering SNP A and SNP B jointly
beyond the phenotype information that would be gained by considering
SNPs A and B independently.

Thus, each edge in a GAIN represents the increase in information
about the phenotype achieved by considering the two SNPs jointly
compared to the expected increase in information with the assumption of
independence between the SNPs. We emphasize that a connection between
SNPs in a GAIN is specific to the given phenotype because it measures
the correlation between two SNPs that influences association with the
phenotype. The network can be exported to Cytoscape or visualized
interactively within the GAIN tool.

GAIN can be combined with SNPrank for a powerful analysis engine. We
have a tutorial
describing the steps of the analysis as well as the dependencies
required.

There are currently two implementations of GAIN available.

• a command-line tool written in Python, hosted on Github
• an older Java-based GUI version of GAIN, hosted on Google Code