Seralogix Mechanistic Model Interrogation and
Discovery

The resulting Bayesian network models that are trained and scored provide considerable insight into the gene-to-gene or protein-to-protein relationships. Interrogation of these models provides valuable insight into which biological function and genes are most likely governed by the treatment or disease being studied. An intermediate (and often iterative) step requires the involvement of the research biologist/scientist to examine the resulting models for clues that help explain the host response. Working with the Seralogix's bioinformaticians, the interpretation of the initial results help select groups of pathways and gene ontology categories that are of high interest. These high interest biological processes form the initial building blocks for creating a biological system model.

Figure 1 shows a typical network model for the MAPK signaling pathway. Figure 2 shows the DBGGA gene scores as a heatmap for the MAPK pathway. Gene/protein nodes on the network model (Fig 1) have the same color coding as on the gene score heatmap. Orange highlighted nodes on the model are those defined as mechanistic genes that surpass a user selected threshold which for this cases is |Bayesian z-score| > 2.24. The model can be visually inspected (static or animated) at any selected time point to view the status of gene/protein expression and correlation between gene relationships. Arcs having thicker lines in the network (Fig. 1) have high positive or negative correlations. A selected set of highly correlated arcs for this model is shown in Figure 3. Interpretation of the arcs and state of gene perturbations provide evidence of pathway points of disruptions or alterations due to the experimental condition.

Such interrogation guides the biologist to identify potential malfunctions and is especially powerful when employed for comparative analysis between multiple experimental conditions. For example, arcs that are reversed in correlation between two conditions may suggest likely candidate genes that are being manipulated as a result of the stimuli/conditions being compared.

return to pipeline
Figure 1. MAPK Pathway Ready for Interrogation

A combination of visualization aids and methods for interrogation of a network model is necessary to hypothesize genetic mechanisms and identify targets of intervention or mechanisms of diseases. For each experimental condition, the trained network model can be interrogated for mechanistic genes across all time points and for the strength (correlation) of connections between interconnecting genes.

Figure 2. Pathway Gene Perturbation Profiling

Each pathway model can produce a heatmap of candidate mechanistic gene scores (at any selected threshold) by time point. The network model highlights these genes (orange rings) and shows the state of expression for all the neighboring genes. For the heatmaps and gene nodes in the network the up regulated genes and their score magnitudes are shown as a gradient color from light to bright red and the down regulated genes are from light to bright green.

MAPK Pathway Correlation between Gene/Proteins

The strength of the relationships between genes (their correlation) is indicated by the thickness of the connecting edges in the model of Figure 1 with dark brown indicating positive correlation and dark green negative correlation.