The hydroisomerization of long-chain alkanes is a key catalytic process for producing high-quality fuels and lubricants, where zeolites play a central role due to their shape-selective properties. However, the vast diversity of zeolite frameworks and alkane isomers poses significant challenges for experimental characterization and predictive modeling. This dissertation develops a comprehensive multiscale modeling framework that integrates adsorption thermodynamics, reaction equilibrium modeling, and machine learning to understand and predict shape-selective effects in zeolite-catalyzed hydroisomerization. A central contribution is the development of the Segregated Explicit Isotherm (SEI) model, which captures adsorbate-size-dependent behavior in heterogeneous adsorbents and enables efficient prediction of multicomponent adsorption. Implemented in the open-source software RUPTURA, this approach supports breakthrough curve simulations, mixture adsorption predictions, and isotherm fitting, providing practical tools for process modeling. To quantify zeolite-induced selectivity at chemical equilibrium, a reaction equilibrium framework was established that combines gas-phase thermochemical properties with zeolite adsorption behavior. Accurate enthalpies, Gibbs free energies, and entropies for long-chain alkanes were obtained using a second-order group contribution linear regression model, which outperforms first-order group methods and achieves sub–chemical-accuracy agreement with experimental and reference data. This model was further extended to compute entropies and exergy destruction, offering insights into the second-law efficiency of hydroisomerization processes. Since adsorption governs equilibrium selectivity in narrow-pore zeolites, special emphasis was placed on Henry coefficients. Molecular simulations provided benchmark adsorption data, while descriptor- and graph-based machine learning models (including TabPFN and directed message passing neural networks) were developed for predictive screening. These models accurately capture adsorption trends for linear and moderately branched alkanes, while active learning strategies improved performance for challenging cases with strong activity cliffs. By combining predicted Henry coefficients with gas-phase thermochemical data, this thesis demonstrates reliable computation of reaction equilibrium distributions for long-chain alkanes in diverse zeolite frameworks. This integrated framework presented here provides both mechanistic understanding and predictive capability, bridging the gap between molecular-level interactions and process-scale performance. It highlights how pore topology, thermodynamics, and molecular structure jointly govern selectivity, and it delivers computational tools that can guide catalyst design, kinetic modeling, and process optimization in hydroisomerization and related hydrocarbon upgrading applications.