Fig. 1

The NPA workflow. The gene expression data are used to estimate the treatment effect for each gene. The (log₂) fold-changes and the associated t-statistics are required (a). By combining the gene expression fold-changes and a two-layer causal network (b), its perturbation is quantified and assessed for its significance by combining two specificity statistics and the fold-change standard deviations (c). Several NPAs can be summarized into a holistic quantity describing the overall biological response, called the Biological Impact Factor (BIF) (d), which can be used as a toxicological index for comparing various treatments. If an individual network reaches significance, the leading node analysis (e) enables the identification of the key biological entities involved in its response to ease the interpretation