GAVA

GAVA is a variant association algorithm, based on the VAAST algorithm. http://www.ncbi.nlm.nih.gov/pubmed/21700766

Computes a p-value for a genomic feature (typically a single gene) using a randomization test, where the affection status of the subjects is permuted and the test statistic computed using a composite likelihood ratio test as described in the VAAST paper.

Note: This is not an exact implementation and for example we have a bailout parameter to avoid spending too much computing time on uninteresting features.

Usage

We need to first set the case and control lists, the disease model (regular, dominant, or recessive), and other parameters. Then the procedure is as follows: * Before processing a feature, call initializeGroup() to clear the counts. * Iterate through the data for the feature and for each variant found for some subject, add the call counts using addFeature(). * Calculate the p-value (and a few other values) using calculateValues(). * Repeat the process for the next feature.

gor ... | GAVA maxiterations -casefile <filename> -ctrlfile <filename> [ attributes ]

or:

gor ... | GAVA maxiterations -caselist <subjects> -ctrllist <subjects> [ attributes ]

Options

-recessive	Use recessive model, i.e. only pick 2 variants per subject.
-dominant	Use dominant model, i.e. only pick 1 variant per subject.
-casepene number	The case penetrance (locus heterogeneity).
-ctrlpene number	The ctrl penetrance.
-noMaxAlleleCounts	No max allele count for | each case in dominant/recessive models.
-protective	Include protective alleles.
-grouping number	Collapsing threshold for rare variants (default is 5).
-bailout number	Bail out of the randomization with this number of hits (default is 10).
-usePhase	Use the phase (in recessive modeling).
-maxAf number	Upper threshold for the estimated control allele frequency.

Warning

casepene and ctrlpene specify the number of cases and ctrls which DO NOT have to comply with the inheritance model.