Pearson/Spearman correlation and types of data in HAllA vs Maaslin2


I have several questions:

  1. In which cases can I use Pearson correlation instead of Spearman in HAllA when analysing 16S and metabolomics data? Metabolomics and 16S data are both continuous data but frequently non-normally distributed and zero-inflated, are there situations in which Pearson is applicable? What about comparing 16S/metabolomics to normally distributed clinical data?

  2. From what I understood, both Maaslin2 and HAllA can be used to analyze correlations between 16S/metabolomics data and clinical data. How do these two methods compare? How does the amount of variables analyzed and the heterogeneity impact the analysis? How do the methods handle colinearity? Is it better to “stick to one method”, for example when using HAllA for the 16S/metabolomics associations to also use it for the clinical associations? And if Maaslin2 is superior, which settings (normalization, transformation, analysis method) can I use in Maaslin2 on the omics data that I am using in HAllA for other analyses to have a comparable result?

Many thanks in advance, sorry for the many questions :slight_smile: