Our secondary analyses revealed a decrease in correlations for Age and Smoking, but an increase in correlation for BMI. Finally, our framework can also be used for prediction, with an accuracy three-fold higher than models based on abundance only.
Feel free to reach out if you have questions!

Overall we highlight strong environmental effects on age, sex, smoking, BMI and to a lesser extent appendicectomy, specific food, NutriNet factor, socio-demographics and cholesterol levels.

We ran a high level multivariate screening of 80 environmental factors at the Family, Genus and Species levels, followed by univariate secondary analyses to further investigate the local changes.

MANOCCA looks at the effect of a predictor on the variations of correlations (co-abundances for microbiome). In the context of microbiology, this can be seen as the presence/absence of a resource which can affect the inter-species relationships and interactions.

Full story available here: www.biorxiv.org/content/10.1...

The method
1. allows to account for complex pleiotropic patterns through its prior,
2. detects more credible sets than univariate SuSiE and its competitor,
3. detects credible sets more often overlapping with regulatory elements than univariate SuSiE.

mvSuSiE builds upon the SuSiE (Sum of Single-Effects) model to leverage statistical evidence across multiple traits.