A New Approach to Analysis of Clinical Data and Prognostication for Patients with Hepatocellular Carcinoma, Based Upon a Network Phenotyping Strategy (NPS) Computational Method.

Abstract

Objectives: There is a multi-component nature of the influences on HCC progression but integrating them has been difficult. Network phenotyping strategy (NPS) integrates all multi-component relationship facets of HCC progression and aims to lead to a new way of understanding human HCC biology.

Methods: We converted baseline patient demographics, tumor characteristics, blood hematology and liver function test results, consisting of values of 17 standard clinical variables, collected time-coherently at the index visit, into a graph-theoretical data representation.

Results: These data were analyzed by NPS, which processes the patient parameter values together with their complete relationships network. NPS identified 25 disease-progression ordered HCC phenotypes. Clinically relevant NPS results are a) Portal vein thrombosis incidence during HCC progression stratified into 5 narrow ranges; b) NPS identified patients according to aggressive, slow and intermediate tumor growth sub-types; c) Personalized prognostication of mortality was achieved by the 25 NPS phenotypes, independently optimized for respective phenotype sub-cohorts.

Conclusion: The NPS results were implemented as an internet application (https://apkatos.github.io/webpage_nps), where input of 17 clinical parameters provides the patient phenotype, phenotype-characteristic average mortality and personal survival estimate.