Background: Alzheimer’s disease is the most common type of dementia, and currently the biggest financial burden for health care providers in developed countries. γ-secretase is a large membrane embedded protease and a major target in drug development efforts based on over 250 mutations in its structure that can cause the disease. We used all-atom and coarse-grained molecular dynamic studies to describe the structural changes in the catalytic cycle of γ-secretase. Results: Adaptive Poisson-Boltzmann calculations showed that γ-secretase is a highly polarized enzyme that can attract N-terminal and C-terminal parts of its protein substrate. The presenilin subunit of γ-secretase has a cluster of positive residues that can stabilize the negatively charged C-terminal of the catalytic intermediates. The nicastrin subunit can close above the catalytic site like a lid or rotate above the enzyme. The 6 Å distance between the active site aspartates showed that the enzyme structure from the Protein Data Bank represents an inactive enzyme. In silico mutations showed that 4.5 Å distance between the active site aspartates is optimal for catalysis. Asparagines in place of the active site aspartates can be used to mimic the active enzyme in structural studies of substrate binding steps. The cholesterol-rich membrane supports the optimal distance between the active site aspartates, what is consistent with experimental results which showed that cholesterol is crucial for the enzyme activity. Disease causing mutation G384A can increase the distance between the active site aspartates, what can explain experimentally observed pathogenic decrease in the activity of mutated enzyme. Conclusion: Molecular dynamic studies can be used for the development of drug candidates that can specifically target changes in γ-secretase structure that support the pathogenic changes in the enzyme activity.