Developing chemical bonding models in clusters is one of the most challenging tasks of modern theoretical chemistry. There are two reasons for this. The first one is that clusters are relatively new objects in chemistry and have been extensively studied since the middle of the 20th century. The second reason is that clusters require high-level quantum-chemical calculations; while for many classical molecules their geometry and properties can be reasonably predicted by simpler methods. The aim of this dissertation was to study doped clusters and explain their chemical bonding. The research was focused on three classes of compounds: aluminum clusters doped with one nitrogen atom, planar compounds with hypercoordinate central atom, partially mixed carbon-boron clusters, and transition metal clusters. The geometry of the two latter classes of compounds was explained using the concept of aromaticity, previously developed in our group. Also the Coalescence Kick Method for finding global minima structure and low-lying isomers was implemented, tested, and applied to the considered cluster systems. Tests showed that the Kick Method works faster than other methods and provides reliable results. It finds global minima even for such large clusters as B17- and B19- in reasonable time.