Understanding the function of biological systems and related diseases requires insights into molecular structural changes and dynamics across multiple length scales, which are often challenging to capture experimentally. This dissertation focuses on the development of multiscale computational simulation frameworks. At the mesoscale, the structure of the human gut mucus network plays a critical role in governing oral drug nanoparticle diffusion and bacterial invasion. In this work, a coarse-grained molecular dynamics (MD) model was developed for the human gut mucus system, capturing the diversity of mucin molecules and the complexity of the biopolymer network. This model accurately characterizes the diffusion coefficients of drug nanoparticles in mucus, consistent with both microscopy measurements and theoretical predictions. Feature design and machine learning methods were employed to analyze nanoparticle diffusion trajectories, revealing how the diffusion modes of spherical particles depend on timescale and particle size. This platform enables the rational design of drug nanoparticle geometry tailored to desired diffusion behaviors and provides a tool for investigating how bacterial activity-induced structural changes in mucus impact disease. At the cellular level, we investigated engineered living concrete as a model system, with bacterial dynamics serving as a key mechanism for tuning the structure and mechanical properties of the material. This work presents a novel agent-based computational model to investigate the dynamics of bacterial enzyme secretion and the catalysis of microbially induced calcite precipitation (MICP), a bacterial process used in reinforcing the mechanical properties of engineered living concrete. This model enables quantitative characterization of how factors such as chemical concentration, bacterial density, and porous structure affect the morphology of calcite precipitation. The approach paves the way for studying the mechanical properties of engineered living materials from the perspectives of bacterial dynamics and material geometry. Finally, integrating the advanced multiscale modeling and machine learning methods, this work designed a physics-informed machine-learned MD potential, which was trained for detailed investigations into the thermal properties, structures, and phases of complex metallic systems.