How many phage species remain undiscovered? Species sampling approaches to inform phage discovery

The emergence of antimicrobial resistant bacteria has been identified as one of the most serious public health and development threats for the near future. The use of bacteriophages (phages) is a promising solution for the sustainable control of these pathogens. Phages are natural viral predators of bacterial pathogens. However, due to the variability and adaptability of bacteria, developing effective and sustainable phage treatments requires drawing from a wide variety of different phage species. This study applies specialised mathematical and computational estimation approaches to the problem of sampling and discovering species of phages in microbiological communities. Our goal is to predict how many new phage species will be discovered in future samples based on species recorded in a phage sequence database for eight common bacterial hosts. While internal validation shows that current estimators yield low prediction errors for how many new species can be found when sampling more phages, these critically assume that the underlying mechanisms of phage sampling and database inclusion remain similar across time - an assumption put into question by our analysis for all but the host genus Mycobacterium . From this, we discuss how our results can still inform future sampling strategies - both through directly predicting the number of additional species, or by detecting changes in sampling and database inclusion. Our results have the potential to inform and optimise the hunt for and isolation of novel of phages from the the natural environment. Highlights Species abundance distribution of phages infecting seven of eight analysed host bacterial genera collected in phage databases changed significantly over time. Maintaining current sampling strategies for phages of host Mycobacteriut may lead to diminishing returns in the number of new phages found in additional samples. Both non-parametric and parametric estimators of the numbers of unseen species show comparable, moderate errors when predicting a random subset from the remainder of hostspecific phage sets from the INPHARED database