PATHiWAYS tool performs an analysis of the impact of a determined condition in the stimulus-response subpathways of selected pathways. The input data is a normalized dataset of a case/control experiment and an experimental design indicating which samples belong to each condition. At the moment, PATHiWAYS tool accepts five types of Affymetrix microarray platforms covering two species: human (Homo sapiens) and mouse (Mus musculus):

• Affymetrix Human Genome U133 Plus 2.0 Array
• Affymetrix Human Genome U133A Array
• Affymetrix Human Genome U133A 2.0 Array
• Affymetrix Mouse Genome 430 2.0 Array
• Affymetrix Mouse Gene 1.0 ST Array

PATHiWAYS tool converts this probe set expression data in activation probability of each gene in the pathway network and then propagate this gene probability through the network obtaining the activation probability of each stimulus-response subpathways of the pathway. Finally, PATHiWAYS performs a multilevel analysis in order to find the significantly differentially activated stimulus-response subpathways between two conditions. As a result, PATHiWAYS provide the numerical and graphical results of this analysis and a representation of this results in the pathway network for a better understanding of the functional consequences of the differential activation of the subpathways.

We selected signaling pathways because they describes a group of molecules in a cell that work together to control one or more cell functions, such as cell division or cell death. After the first molecule in a pathway receives a signal, it activates another molecule. This process is repeated until the last molecule is activated and the cell function is carried out.

As a background PATHiWAYS tool has stored two types of information:

1. Modelled KEGG signalling pathways networks.
2. Modelled probability distributions of all probe sets of each platform.

For the first type of information, we model 27 signaling pathways for Human (Homo sapiens) and 18 for Mouse (Mus musculus) of KEGG pathways database. In this pathway networks, nodes are representing one or more genes and edges are representing the relationships between genes. The modelling of these pathways follows two principal steps:

1. Convert all the relationships between genes in two types of relations: activations and inhibitions.
2. Group all the genes that work together to transmit the signal through the network.

For the second type of information, we model the distribution of the expression of each probe set as a mixture of probability distribution. This distribution allows the calculation of the activation probability of each gene depending on its expression value in a microarray experiment.

PATHiWAYS analysis performs four principal steps:

1. Inferring the probability of activation of each node in the pathway based on the level of expression and the mixture for each probe set belonging to that node. To summarize the information obtained of each probe set on the information used in the node, we can use one of the following functions: mean, median, maximum, minimum or different percentiles. This is the parameter called summ in other parameters section in PATHiWAYS form.
2. Estimating the probability of activation of each stimulus-response subpathway in the pathway for each sample based of the activation probability of each node and the relationships between them.
3. Performing a multilevel analysis in order to asses the differential activation probability of every stimulus-response subpathway in each pathway, given the status of the sample.
4. Represents the results in a pathway context.