Fig. 7: Associations identified by MOFA in predator-exposed fish among behavioral traits and microbial abundances, for factor 3.

Factor 3 explained 13% of the variance in the behavioral dataset and 1% of variance in the microbiome dataset in predator-exposed fish. Factor 3 was not significantly correlated with population or parasite status, although there was a non-significant weak correlation with population (p > 0.05). A Contribution (weight) of behavioral traits to factor 3. Traits are ranked according to their weight. The higher the absolute weight, the more strongly associated a trait is with the factor. A positive weight indicates the trait has higher levels in samples with positive factor values, while a negative weight indicates the opposite. Behavioral traits are colored by their broader behavioral category. The top five traits contributing to the factor in each direction are labeled. Behavioral trait abbreviations: AT measurement after predator trigger, BT measurement before predator trigger, dist distance, Dtot total distance traveled, dur duration of time spent in the respective tank zone, freq number of times fish visited the respective tank zone, pred measurement relating to the position of the predator, Vang angular velocity, VelBL mean velocity in body length per second. B Contribution (weight) of microbial genera to factor 3, displayed in the same way as for (A). Microbial genera are colored by their putative source and shaped on their potential as a fish pathogen, based on a literature search (Supplementary Table 5). C Example association between factor 3 sample weights vs a top behavioral trait, angular velocity before simulated predator attack (BT). Samples are colored by fish population. A smoothed linear conditional means line is shown in black with confidence intervals in gray. D Example association between factor 3 sample weights vs a top microbial genus, Paracoccus, displayed in the same way as for (C).