The human liver is responsible for essential processes like fat storage or the detoxification. Some liver diseases can trigger growth processes in the liver, disrupting important hepatic function-perfusion processes [1]. To understand the interplay between hepatic perfusion, metabolism and tissue in the hierarchically organized liver structure, we have developed a multicomponent, poro-elastic multiphasic and multiscale function-perfusion model, cf. [2,3], using a multicomponent mixture theory based on the Theory of Porous Media (TPM, see [4]). The multiscale approach considers the different functional units of the liver, the so-called liver lobules, with an anisotropic blood flow via the sinusoids (slender capillaries between the periportal field and the central vein), and the hepatocytes, where the biochemical metabolic reactions take place. On the lobular scale, we consider a tetra-phasic body, composed of a porous solid structure representing healthy tissue, a liquid phase describing the blood, and two solid phases with the ability of growth and depletion representing the fat tissue and the tumor tissue. To describe the influences of the resulting tissue growth, the model is enhanced by a kinematic growth approach using a multiplicative split of the deformation gradient into an elastic and a growth part, dependent on the fat accumulation and tumor development. To describe the metabolic processes as well as the production, utilization and storage of the metabolites on the cellular scale, a bi-scale PDE-ODE approach with embedded coupled ordinary differential equations (ODE) is used. In order to represent realistic conditions of the liver, experimentally or clinically obtained data such as changes in perfusion, material parameters or tissue morphology and geometry are integrated as initial boundary conditions or used for parametrization and validation [5]. Data integration approaches like machine learning are developed for the identification, processing and integration of data. A workflow is designed that directly prepares the model for clinical application by (semi-) automatically processing the data, considering uncertainties, and reducing computation time. REFERENCES 1. Christ, B., …, Ricken, T. et al. [2017], Frontiers in Physiology 8, 906, doi: 10.3389/fphys.2017.00906 2. Ricken, T., et al. [2015], Biomech. Model. Mechanobiol. 14, 515-536, doi: 10.1007/s10237-014-0619-z 3. Ricken, T., and Lambers, L. [2019], GAMM-Mitteilungen 42(4), doi: 10.1002/gam