CorrelationCalculator for Metabolomics Data
CorrelationCalculator is a standalone Java application providing various methods of calculating pairwise correlations among repeatedly measured entities. It is designed for use with quantitative metabolite measurements such as MS data on a set of samples. The workflow allows inspection and/or saving of results at various stages, and the final correlation results can be dynamically imported into MetScape (version 3.1 or higher) as a correlation network.
Download CorrelationCalculator
Download sample input file: aminoacids.csv
Download sample input file: kora_data_240.csv
Input File Format
For this release, input files are required to be in CSV format. Input files should be tables of measurements of metabolites across multiple samples. Both metabolites and samples must be labeled. Samples may be in either rows or columns. The example input file has its samples in rows.
CorrelationCalculator Workflow
After a data file has been imported, users have the option of normalizing
the data using a log transformation and/or autoscaling.
Pearson's correlation coefficients are useful both to get an overview of
correlation structures and as a way of filtering large data sets.
Results may be viewed as a static heatmap in PDF format or exported to
a file or an interactive heatmap viewer such as TreeView.
A slider is provided for filtering data sets based on a Pearson's correlation
threshold. Metabolites with no Pearson's correlations above the threshold
value are excluded from subsequent analyses.
Partial correlations can be calculated on the filtered data using a Debiased Sparse
Partial Correlation (DSPC) algorithm developed by our collaborators George
Michailidis and Sumanta Basu. The DSPC method is particularly useful when
the number of metabolites exceeds the number of samples in the data set.
Partial correlation results can then be viewed interactively in
MetScape as a correlation network.
Cite CorrelationCalculator
Basu S, Duren W, Evans CR, Burant C, Michailidis G, Karnovsky A.
Sparse network modeling and Metscape-based visualization methods for the analysis of large-scale metabolomics data.
Bioinformatics. 2017 Jan 30
