Final Activity Report Summary - GALAGN (Galaxy formation and evolution and the role of AGN) Recent observations and theories support an evolutionary scenario where star formation and accretion activity in galaxies are connected. In this work, we investigated such a link by studying the multi-wavelength properties of active systems, both starburst galaxies, and 'Active galactic nuclei' (AGN) at high-redshifts. Accretion activity and star formation are quantified using mid-infrared (mid-IR) data from the Spitzer Space Telescope and millimetre data from the IRAM 30m Telescope. A population of massive starburst galaxies at high-redshifts and with extreme star formation rates (SFRs) was discovered. Based on the derived large stellar masses, high SFRs and lack of AGN signatures, it is suggested that the mechanism that halts star-formation in powerful starburst galaxies at high-z might not be AGN driven.Two bright millimetre AGNs at z=3.5 were discovered and studied in details. They are characterised by both powerful starburst and AGN activities at similar levels, by outflowing material, large metallicities and stellar masses, heavily absorbed X-ray emission, and large mid-IR/X-ray luminosities. These sources represent the rare transition phase between the starburst and AGN phase predicted by the most recent evolutionary models. Their properties rule out the standard AGN unification model, and favour either a clumpy dust distribution or a scenario with dust in the torus opening angle. A detailed study of the dust properties of obscured AGNs at high-redshifts was also carried out by modelling their IR SEDs and spectra. From the modelling, it emerges that obscuration is not always associated with the torus, but in some cases with cold dust along the line of sight. Based on this study, the decrease in the fraction of obscured AGNs at high-luminosity is confirmed, but more obscured AGNs are found than in optical and X-ray samples at similar luminosities. We find intense star-formation activity in these AGNs, but its contribution to the bolometric luminosity is, on average, <20% of the total bolometric luminosity. By comparing the multi-wavelength properties of different types of obscured AGNs at high-redshifts, we find higher stellar masses in systems with lower AGN power. This trend can be explained by either decreasing Eddington ratios or increasing offsets from the local black hole-bulge mass relationship. In summary, we find that (1) an AGN is not always present in high-z starbursts suggesting that BH growth occurs on shorter timescales than star-formation; (2) the mechanism that halts star-formation in massive starburst galaxies at high-z might not be AGN-driven; and (3) the black hole is in place before the stellar mass has been completely formed or assembled.We have also modelled, for the first time, the infrared spectra of obscured QSOs with torus models providing strong support to the receding torus scenario, demonstrating that obscured AGNs are more common than unobscured AGNs, even at high luminosities and that optical and X-ray selections are biased against these objects, and that these systems are hosted by powerful starburst galaxies. All these studies are presented in four refereed articles, two in the Astrophysical Journal and two in Astronomy & Astrophysics. These studies provide new insights on the star-formation process at high-z, and observables that can be used to test the latest evolutionary theories and better understand the high-redshift universe.