DART

"DART" (Differential Activation of Pathways Tumor Analysis) introduces a groundbreaking network-based methodology for estimating the activation patterns of molecular pathways in tumors. This methodology utilizes model signatures derived from perturbation experiments and structural models of molecular interactions. This innovative approach is designed to address the challenge of understanding the heterogeneity in clinical responses across various tumor types and to aid in developing more effective cancer therapies.

DART is built on the observation that pathway networks inferred from cancer expression data align well with the prior knowledge encapsulated in model signatures. However, they also display a modular structure. Crucially, DART posits that the estimation of pathway activity is intricately linked to this modular structure, enabling a more nuanced interpretation of pathway activation in tumors.

DART's innovative use of Boolean interaction Cox-regression models uncovers non-linear pathway combinations associated with clinical outcomes, offering prognostic models superior to those based on individual pathways. For instance, in ER+ breast cancer, it finds that high MYC and RAS activity in combination predict a significantly worse prognosis than high activity in either pathway alone.