Innovative data mining in life sciences depends on the availability of novel approaches and their freely accessible implementations. As part of the CompLiMet site we are providing an online implementation of Signed Distance Correlation and Partial Distance Correlation – SIDCO.

Distance correlation, a non-parametric approach for correlation analysis, can measure linear and nonlinear, monotonic and polytonic data relationships. In metabolomic and lipidomic datasets, distance correlation can take into consideration the sparse coverage of feature data, the potential for determining non-linear relationships, as well as possibly random network topologies associated with metabolism and inherent to lipidomic and metabolomic datasets. Distance correlation is zero only for fully independent features. The online implementation presented in this publication is available at https://complimet.ca/sidco and provides an easy implementation of distance correlation for any size dataset as well as solution for partial distance correlation, estimating values for direct correlations only. Output provides correlation and corresponding p-values for each feature pair for further analysis. Our supplementary material provides an example of the difference between complete and partial correlation results in metabolomics data analysis. For further details about the algorithm, methodology and its online implementation please check out our publication or SIDCO site.

The figure above shows an outline of SIDCO site providing one-to-one, pairwise correlation for each feature consecutively to all other feature; one-to-all correlation measuring correlation between each feature and all other features combined and finally partial correlation analysis calculating only direct correlations between pairs of features.