A grand challenge in biology is the lack of annotation for much of global sequence space. In particular, while microbes and their viruses are critical to Earth System functioning, surveys of their communities in nature routinely return genes of unknown function, particularly for viruses. This hampers understanding and predictive capability, and reflects a number of database limitations. Here we apply high-resolution environmental metaproteomics to 4 purified viral concentrates from the Tara Oceans Expedition to identify in situ proteins extracted from uncultivated viral particles. Using matched metagenomic databases, we identified 1875 proteins belonging to 549 non-redundant protein clusters that are predominantly of unknown function. These data help identify function for structure-associated proteins in known abundant tailed viruses (e.g., pelagiphages and cyanophages) and validate newly-identified uncultured viral sequences linked to abundant novel hosts. One protein cluster alone comprised 22% of the identified structure-associated proteins, with ~46% of the observed spectra. Structural modeling of this highly abundant viral protein cluster indicates its likely role as a novel, previously unknown capsid. Together these analyses provide much-needed, culture-independent structure-associated protein annotations critical for identifying viral signals across diverse datasets to better elucidate viral roles in nature.