MIMOSA2

What Does MIMOSA2 Do? Description and Terminology

Input Data

Choosing Settings and Running the Analysis

Results

R Package Tutorial

Reference Data

FAQs

Citation and Contact

Running a MIMOSA2 Analysis: Setup and Settings

The MIMOSA2 web application is available at http://elbo-spice.cs.tau.ac.il/shiny/MIMOSA2shiny/

The main web application page is an interface for uploading data and selecting options for your MIMOSA2 analysis. The various sections and options are described below:

Microbiome Data

Select your microbiome data format, and upload the corresponding data file. You can provide taxonomic abundances, function abundances, or both, in several formats. See Input Data for more information.

Metabolic Model Options

Select what metabolic database you would like to use to link your microbiome and metabolite data. Several options are available:

If you have provided a table of 16S rRNA ASVs, you can select how strictly these sequences are aligned against the reference sequences. The default is a relatively strict minimum threshold (99%). Mappings are pre-computed for reference OTU sequences at a 97% threshold.

Finally, advanced users can optionally provide a file specifying modifications to the network model, including adding custom reactions and filtering specific reference reactions. See Input Data.

Metabolomics Data

Select your metabolite data format, and upload the corresponding data file. Metabolite data can be generated using any platform but should consist of identified metabolite abundances (preferably specified as KEGG Compound IDs)

Select whether the metabolite data should be log transformed. We recommend this option if your metabolite measurements tend to have highly skewed distributions (for example, if your study consists of two sample groups and one has greater variability than the other). You may find it useful to examine the CMP-Metabolite comparison results with and without log transformation.

Algorithm Settings

Select whether to compare metabolic potential (CMP) and metabolite levels using ordinary least-squares regression (OLS) or rank-based regression. We generally recommend rank-based regression as it can detect metabolite relationships more robustly and sensitively across a wider variety of data distributions. However, for rank-based regression, the contributions of individual taxa to the model fit are calculated using a permutation-based approach, which greatly increases the analysis runtime. Therefore, when this option is selected, MIMOSA2 will only calculate taxonomic contributors for metabolites with a model p-value less than 0.1 (rather than all metabolites). You also have the option to skip the contribution analysis and just compare metabolites with metabolic potential.

Rank-based regression is provided using the Rfit package in R. The general idea is to find a regression solution that minimizes a function of both the rank and the size of the model residuals (instead of the sum of squared errors in OLS). You can read more about the statistical approach here. Other non-linear models may be provided as options in the future.

Run MIMOSA

Once you have selected all options, push this button to run your analysis. Analysis runtime depends on the numbers of samples, microbiome taxa, and metabolites in your dataset. Analyses using OLS regression typically only take a minute or two; analyses using rank-based regression may take 5-10 minutes.

If your connection to the server is lost mid-analysis, it may be due to server load - please try again later. If the analysis finishes but then disconnects while examining the results, you can still download a zipped directory of your results from the server using the provided link.