This page lists the terms that Fisher's exact test proposes as possible explanations for the genes in the active set.
Fisher's exact test is simpler than the alternative (model-based MCMC-sampling) analysis implemented by this app, and it has pros and cons relative to that approach. One advantage is that it is very quick: results will show up on this page instantly, as soon as you have entered some gene names on the Data page. A related advantage is that its convergence is guaranteed, so you don't need to worry about whether it's run for long enough, as is sometimes the case with the MCMC analysis.
On the negative side, Fisher's test ignores interactions between the terms, unlike the model-based approach, which allows multiple terms to "collaborate" in explaining the data. Fisher's test tends to identify a lot of terms as being enriched, and some of the interesting signal can get lost in the noise.
The model has three parameters: π, α, and β. These specify the probabilities of three types of random event, namely: term activation (π); mislabeling of inactive genes as active, i.e. false positives (α); and mislabeling of active genes as inactive, i.e. false negatives (β).
A beta distribution prior is specified for each of the three parameters. Each beta prior has two hyperparameters, representing pseudocounts for the number of times the corresponding random events did or didn't occur.
Parameter π is the probability that a term will be active, thereby activating all its associated genes.
The most likely value for π under the prior.
The total number of pseudocounts, determining how much influence the prior has relative to the model.
Pseudocounts
Parameter α is the probability that a gene will appear in the active set, despite no associated active terms.
The most likely value for α under the prior.
The total number of pseudocounts, determining how much influence the prior has relative to the model.
Pseudocounts
Parameter β is the probability that a gene will not appear in the active set, despite one (or more) associated active terms.
The most likely value for β under the prior.
The total number of pseudocounts, determining how much influence the prior has relative to the model.
Pseudocounts
The number of samples per term that the sampler is allowed to run before statistics-gathering commences.
The target number of samples per term that the sampler will be run while statistics are gathered.