Biomarkers for capturing disease pathology as molecular process hyperstructure

Abstract Background Precision of drugs in clinical development but also of approved treatment sees limits, documented as attrition in clinical stage drug testing and suboptimal number needed to treat in clinical practice. Precision medicine aims at approaching a causal relation of disease pathology, treatment mechanism of action and clinical outcome. The instance linking pathology, clinical phenotype and drug response is disease characteristics amenable for quantitation, including established clinical phenotyping parameters and upcoming molecular profiling and biomarkers. Molecular biomarkers situated at the interface of pathology-specific molecular process architecture and drug mechanism of action promise capturing aspects allowing assessment of treatment response. Results Approximating a set of 1,008 disease terms as pathology molecular networks provides 3,860 molecular processes involving 4,602 protein coding genes. Assembling this process set in a hierarchical cluster using mean shortest paths among processes as distance measure allows representation of molecular processes in cumulative aggregation. This procedure transforms human disease pathology into a static instance of a molecular process hyperstructure involving 1,340 aggregate levels in a molecular architecture. The hyperstructure allows evaluating molecular biomarker candidates at different levels of molecular process aggregation in terms of biomarker-specific entropies. Interpretation as information content reflects the capacity of a biomarker for sensing molecular process configuration. Deriving entropies across aggregation levels for a reference set of 1,502 biomarker candidates identifies significant spread in information content of individual biomarkers. Exemplified on biomarker panels holding evidence for prognostic capacity and factors serving as drug targets from selected chronic diseases, biomarker entropies allow interpretation in terms of sensitivity for capturing process context and specificity for informing on the status of individual processes afflicted with a given pathology. Conclusions High entropy biomarkers provide candidate molecular proxies for clinical phenotyping parameters, and low entropy biomarkers add information on specifics of disease pathology. Combining high and low entropy biomarkers in panels may offer relevant resolution of molecular process configurations for improving patient stratification with respect to minimizing variance in drug response.