Proteins are biological macromolecules composed of amino acids linked by peptide bonds. Their three-dimensional (3D) structures are still challenging to determine and the number of proteins with resolved tertiary structures is rather small compared to the number of known protein sequences. The 3D structures of proteins are essential for understanding their function, and thus biological processes orchestrating health and diseases. The 3D protein structure allows us to identify "binding pockets" and functionally relevant regions of the protein. Nowadays innovative approaches have been developed for fast determination of protein conformations. These include computer algorithms that predict the 3D structure of the protein from its polypeptide primary sequence. In this thesis, we use AlphaFold 2, an open-source software that uses available protein datasets and artificial intelligence (AI), to predict the 3D structure of proteins. In this study, AlphaFold 2 structure models were analyzed for randomly generated amino acid sequences and for well-known industrial biocatalysts halohydrin dehalogenases HheC and HheA. The random sequences were generated by the tool RandSeq, while the FASTA inputs for HheA and HheC were formed from the crystal structures 1ZMO and 1ZMT, respectively, downloaded from the Protein Data Bank. The AlphaFold 2 conformations were analyzed using PyMOL and ChimeraX visualization software. While AlphaFold 2 could not reliably predict the structures of random sequences, as expected, the structures of the enzymes HheA and HheC in their monomeric and tetrameric states were predicted with high reliability. However, structural peculiarity like the entry of the C-terminal tail into the diagonal subunit of the HheC tetramer was not predicted. This study shows that AlphaFold 2 structures can be good starting conformations for molecular dynamics simulations while their use for molecular docking calculations should be taken with caution.