Combination of statistical methods for easy analysis and classification of virus–host environmental dynamics in the saltern of Sfax, Tunisia

Investigations of virus–host interactions depend heavily either on their cultivation or on metagenomics. In this study, we combined complementary statistical methods and culture-independent methods to gain deeper insight into virus–host interactions in the Sfax saltern, Tunisia. Samples were collected and the data were statistically evaluated. Hierarchical cluster analysis classified the ponds into two main groups with a sharp cutoff: type A with a salinity value < 24% and type B with salinity > 24%. Redundancy analysis demonstrated a strong correlation between viral abundance, virus-to-prokaryote ratio, and archaea in type B ponds. Head-tailed viruses with medium–small genome size were often predicted to preferentially infect the most abundant hosts in high-salinity ponds, Salinibacter and Haloquadratum , which are both strongly correlated with high Mg 2+ concentrations. Based on the redundancy analysis, type A ponds were defined by relatively high Ca 2+ content and were dominated by spherical viruses, with the largest genomes infecting either the potential bacterial host Halopeptonella or the archaeal host Halogeometricum . Interestingly, the low-salinity pond M1Oct09 displayed a special viral community pattern. The receiver operating characteristic curve clarified the criteria that could not discriminate between pond types A and B, including Na + , pH, and bacterial abundance. Parameters including the viral and archaeal diversity index and the virus morphotype showed a correlation with salinity but could not be used to discriminate between these types. Combining the complementary statistical and culture-independent methods could serve as suitable evaluation tools for providing clear and easily understandable information on virus–host interactions.