Pairwise learning for predicting pollination interactions based on traits and phylogeny

Prediction of Klebsiella phage-host specificity at the strain level

Phages are increasingly considered promising alternatives to target drug-resistant bacterial pathogens. However, their often-narrow host range can make it challenging to find matching phages against bacteria of interest. Current computational tools do not accurately predict interactions at the strain level in a way that is relevant and properly evaluated for practical use. We present PhageHostLearn, a machine learning system that predicts strain-level interactions between receptor-binding proteins and bacterial receptors for Klebsiella phage-bacteria pairs. We evaluate this system both in silico and in the laboratory, in the clinically relevant setting of finding matching phages against bacterial strains. PhageHostLearn reaches a cross-validated ROC AUC of up to 81.8% in silico and maintains this performance in laboratory validation. Our approach provides a framework for developing and evaluating phage-host prediction methods that are useful in practice, which we believe to be a meaningful contribution to the machine-learning-guided development of phage therapeutics and diagnostics.

Predicting plant-pollinator interactions: concepts, methods, and challenges

Plant-pollinator interactions are ecologically and economically important, and, as a result, their prediction is a crucial theoretical and applied goal for ecologists. Although various analytical methods are available, we still have a limited ability to predict plant-pollinator interactions. The predictive ability of different plant-pollinator interaction models depends on the specific definitions used to conceptualize and quantify species attributes (e.g., morphological traits), sampling effects (e.g., detection probabilities), and data resolution and availability. Progress in the study of plant-pollinator interactions requires conceptual and methodological advances concerning the mechanisms and species attributes governing interactions as well as improved modeling approaches to predict interactions. Current methods to predict plant-pollinator interactions present ample opportunities for improvement and spark new horizons for basic and applied research.

Nesting biology and nest structure of the exotic bee Megachile sculpturalis

From the 1990s, the Southeast Asia native giant resin bee Megachile sculpturalis (Smith, 1853) was introduced first to North America, and then to many countries in Europe. Despite increasing studies on its invasive potential and geographical expansion, information on nesting behaviour of this species is still extremely scarce. To increase knowledge on the nesting biology of M. sculpturalis, we studied multiple aspects of nesting and pollen provisioning in three consecutive years in artificial nests in Bologna, Italy. We observed 166 bees visiting nests, and followed individual nesting behaviour and success of 41 adult females. We measured cavity diameter in 552 nests and characterised the structure in 100 of them. More than 95% of nest diameters ranged between 0.6 and 1.2 cm, overlapping with several sympatric species of cavity-nesting hymenopterans in the study area. Most nests had a first chamber from the entrance of variable length without brood, followed by an average of about two brood cells with a mean length of 2.85 ± 0.13 cm each. The pollen stored in brood cells was almost monofloral, belonging to the ornamental plant Styphnolobium japonicum (L.) Schott. We estimated that a single female should visit ≈180 flowers to collect enough pollen for a single brood cell. These results fill knowledge gaps on the nesting biology and nest structure of the exotic M. sculpturalis, and they are discussed in relation to possible competition with native bees for nesting sites and foraging resources.

Network embedding unveils the hidden interactions in the mammalian virome

Predicting host-virus interactions is fundamentally a network science problem. We develop a method for bipartite network prediction that combines a recommender system (linear filtering) with an imputation algorithm based on low-rank graph embedding. We test this method by applying it to a global database of mammal-virus interactions and thus show that it makes biologically plausible predictions that are robust to data biases. We find that the mammalian virome is under-characterized anywhere in the world. We suggest that future virus discovery efforts could prioritize the Amazon Basin (for its unique coevolutionary assemblages) and sub-Saharan Africa (for its poorly characterized zoonotic reservoirs). Graph embedding of the imputed network improves predictions of human infection from viral genome features, providing a shortlist of priorities for laboratory studies and surveillance. Overall, our study indicates that the global structure of the mammal-virus network contains a large amount of information that is recoverable, and this provides new insights into fundamental biology and disease emergence.