Mechanistic models offer a realistic framework to understand how signal transduction depends on changes in the expression of genes involved in signaling pathway circuits (see Hidalgo et al., 2017). Interestingly, cell signaling circuits trigger relevant cell functionalities (e.g., proliferation, cell death, metabolic changes, etc.), which virtually allows to estimate not only signaling activity profiles but also a cell functional activity profiles, from transcriptional profiles. Moreover, a mechanistic model allows predicting the effect of interventions (gene knock-outs, drugs, etc.) over a given sample, opening the door to multiple applications for research or therapeutic purposes.

The HiPathia web server integrates four different pathway analysis modules:

• Differential signaling provides an estimation of significant cell signaling activity changes across different conditions.
• Prediction allows you to train a prediction test and test it with different data.
• Perturbation effect is an interactive working environment to simulate the effect of different interventions (e.g. knock-out, over-expression, etc.) over the activity of signaling circuits in the pathways, as well as their potential functional consequences in the cell.
• Variant interpreter provides an estimation on the potential impact of genomic variation on cell signaling and ultimately on cell functionality.

This module provides an estimation of the significant cell signaling activity changes across different conditions. To achieve so, signal value activities for each signaling circuit defined for each of the sample studied are estimated. Then, these signaling activity profiles can be compared in a testing framework as:

• Between two groups, to check which signaling circuits present a significant change among the two grups compared, or
• With a continuous value (e.g., the level of a metabolite, etc.), providing a correlation of each signaling path with this variable.

See the Differential signaling module.

HiPathia allows you to train, download and test a prediction model for your dataset. The model can be trained either to:

• Discriminate between two different groups of samples, or
• Predict the value of an unknown variable.

In order to check how to use these options please see Prediction.

This module offers an interactive working environment to simulate the effect of different interventions (e.g. knock-out, over-expression, etc.) over the activity of signaling circuits in the pathways, as well as their potential functional consequences in the cell. Given a specific condition, corresponding to a specific gene expression profile, knock-outs, knock-downs or different types of inhibitions are simulated by reducing or setting to 0 the expression level of the gene or interest. Conversely, overexpressions or agonistic interventions can be simulated by increasing its gene expression level. Then, the new condition with the simulated interventions is compared to the original one and the circuits affected are reported. In this way a holistic view of the ultimate effects across all the signaling pathways can be obtained. The simulation of the intervention can be carried out in different ways:

• Over one or several genes, or
• Simulating the effect of one or several drugs, by simulating the effect of the drug on the corresponding genes targeted, or
• A combination of the two previous scenarios

See the Perturbation effect module.

This module provides an estimation of the potential impact of genomic variation over cell signaling and consequently on cell functionality. The tool predicts the possible effect that gene disruptions produced by LoF (Loss of Function) variants can have over signaling circuits as well as the cellular functions triggered by them. This application of the HiPathia mechanistic model would be a specific case of the perturbation effect analysis. A knock-down simulation is carried out on genes harboring LoF variants in a condition that represent a normal tissue. Then a perturbation effect analysis is carried out by comparing the condition with the simulated KOs with the normal condition. Normal conditions are taken from the repository of gene expression profiles of normal tissues available in the GTEx database or, alternatively, a gene expression profile can be provided by the user. Then, the signaling circuits potentially deregulated by the disruption of the gene/s provided by the user are detected and visualized.

See the Variant interpreter module.