PatientProfiler: A network-based approach to personalized medicine

Deciphering the mechanisms underlying reprogramming in cancer cells remains a crucial challenge in systems oncology, as it is key to improving patient diagnosis and treatment. For this reason, comprehensive and patient-specific multi-omic characterization of tumor specimens has become increasingly common in clinical practice. The main shortcoming is the absence of robust computational frameworks to integrate and interpret the available multi-dimensional data, to drive translational solutions. To fill this gap, we developed PatientProfiler, a computational workflow that leverages causal interaction data to address how the genetic and molecular background of individual patients contributes in establishing a malignant phenotype. PatientProfiler allows multi-omic data analysis and standardization, generation of patient-specific mechanistic models of signal transduction, and extraction of network-based prognostic biomarkers. We successfully benchmarked the tool to proteogenomic and clinical data derived from 122 biopsies of treatment-naïve breast cancer, available at the CPTAC portal. We identified patient-specific mechanistic models that recapitulate dysregulated signaling pathways. In-depth topological exploration of these networks allowed us to define seven subgroups of patients, associated with unique transcriptomic signatures and distinct prognostic values.