This thesis addresses advances in the field of computational Nuclear Magnetic Resonance (NMR) with two specific objectives: 1) developing an approach enabling the direct probing of intramolecular electronic effects on molecular properties; 2) assessing and improving the accuracy of NMR chemical shift computations for large molecules and large–scale collections of molecules. We first introduce a quantum chemical formalism, which combines the Block– Localized Wave function (BLW) approach with the Individual Gauge for Localized Orbitals (IGLO) methodology for computing NMR properties. The BLW–IGLO method enables the direct probing of the effects of electron delocalization on magnetic response properties using both standard (delocalized) molecular structures and those with "non–interacting" (localized) double bonds. Illustrative examples of the methodology clarify abnormal chemical shifts and solve long–standing problems in organic chemistry. The BLW procedure further serves to identify the remarkable structure– property relationships connecting the activation barrier of the Cope rearrangement and the ground state properties of various derivatives of semibullvalene. Such a correlation provides a straightforward access to kinetic parameters without actually performing the demanding transition state search. It could be utilized as the last step of a hierarchical screening to assess the kinetic feasibility and persistence of molecular targets. The second objective of this thesis focuses on large–scale NMR chemical shift computations, which offer the adequate balance between accuracy and feasibility. Simulation and prediction of NMR chemical shifts constitutes a helpful vii tool for synthetic chemists. While numerous software packages allow for the simulation of NMR chemical shifts, they make use of different methodologies for which relative performances and inner working remain unknown to the user. The accuracy of four parameterized and one ab initio methods were evaluated based on the computations of over 650 1H NMR chemical shifts associated with more than 170 organic molecules. On the basis of these results, guidelines and recommendations are provided. The implementation of a QM/MM formalism devised for the computation of NMR chemical shifts of large systems concludes this thesis. In the proposed approach, the single point charges generally used to treat the long–range electrostatic interactions are replaced by smeared charges with a finite width, with the aim of correcting the overpolarization of the QM region caused by the point charge approximation. The performance of the scheme is compared against standard QM/MM and Full QM procedures and suggestions for future work are proposed.