In this thesis, composition and dynamics of the microbial community of the central Adriatic were determined during a one year study on a spatio-temporal scale, with an emphasis on aerobic anoxygenic phototrophs (AAPs). A combination of 16S rRNA metabarcoding and quantitative CARD-FISH approach was used to determine the structure of the total microbial community at finer taxonomic scale. The composition of the AAP community was estimated by pufM gene metabarcoding and IR-FISH method. In order to detect and interpret patterns within microbial communities in relation to the most significant environmental factors, artificial neural network algorithms were implemented. Members of SAR11 clade prevailed in diverse environmental conditions. A general pattern was recorded of total microbial diversity increase at the expense of abundance in oligotrophic conditions, especially with increased salinity. A significant seasonality was detected in average absolute AAP counts (maxima in spring: 2.136 ± 0.081 x \(10^{4}\) cells \(mL^{-1}\), minima in summer: 0.86 x \(10^{4}\) cells \(mL^{-1}\)), IR-FISH groups (Roseobacter clade predominant in autumn, classes Alpha- and Gammaproteobacteria in summer) and pufM dataset (differences in diversity measures). No differences were observed in the spatial nor vertical profiles. The community had a heterogeneous composition with the highest average relative proportion of the Roseobacter clade (37.66%), followed by members of Gammaproteobacteria (35.25%) and Alphaproteobacteria (31.15%). The prevalence of genus Luminiphilus was revealed at all stations, months and depths. An inverse relationship between AAP abundance and diversity was observed in summer, but not for other seasons. High AAP diversity was independent of temperature, salinity and trophic status of the environment. Additionally, a detailed comparison of five DNA isolation protocols and direct PCR was performed on mock and marine community samples as a fundamental step in metabarcoding. No method was closest to the theoretically expected mock community composition. It has been demonstrated that conventional and toxic phenol-based protocol can be replaced by its shorter and safer alternatives based on isopropanol and ethanol precipitation, while direct PCR results have shown advantages of reduced sample manipulation time.