Industrial heterogeneous catalysts are very complex systems. Unraveling their atomic-scale structures and understanding their roles in the catalytic reaction are important challenges. In the present work, we show how Density Functional Theory (DFT) calculations were used to provide an original information about the structure for active sites of complex catalytic systems of industrial relevance, as a function of their environment, to assign spectroscopic observations and to quantify the kinetics of multi-step reactions they can catalyze. Heterogeneous catalysts involved in industrial applications such as refining, petrochemistry, biomass conversion and pollution abatement were considered. In particular, original models of imperfect aluminosilicates (dealuminated zeolites, external surfaces of zeolites, amorphous silica-aluminas) were developed and their Brønsted acidity unraveled. The reactive environment-dependent (hydrogen and hydrocarbons) morphology of subnanometric platinum-based clusters was revealed. Finally, we show how ab initio thermodynamic and kinetic information can be introduced in kinetic models, possibly integrated themselves in Computational Fluid Simulations, to access macroscopic predictions thanks to a multiscale approach. These achievements pave the way for research perspectives oriented towards the building of predictive macroscopic models for reactions of industrial relevance catalyzed by complex aluminosilicates and supported metals.