In the following bioinformatics paper, "Quantifying environmental adaptation of metabolic pathways in metagenomics", Gianoulis et al. employ the use of two tools to detect multivariate relationships between environmental features and microbiomic features:

  • Regularized Canonical Correlation Analysis
  • Discriminative Partition Matching (DPM)

The paper references several statistics papers and books, including a book by R. Wichern, but it unclear which of these is the reference for DPM. A Google search for "Discriminative Partition Matching" pulls up several applied papers, including the Gianoulis paper, but no direct expositions of the method.

Where can I find information on "Discriminative Partition Matching"?


1 Answer 1


I believe the method is by Gianoulis herself. From her thesis, which has been published as a book "Mining Biological Complexity: Cross Integration of Large-Scale Metagenomics, Environmental, and Chemical Datasets"

4.5.6 Discriminative Partition Matching (DPM)

To analyze whether groupings of sites based on similar environmental features also shared functional similarities, we clustered the sites based on their quantitative environmental metadata resulting in two distinct clusters or site-sets. Next, we partitioned the sites in the metabolism matrices (see Figure 4.1A) into the same two site-sets and calculated the mean normalized frequency for each pathway in each site-set (see below for generalized approach). If the means of the pathway frequency between the two site-sets were not significantly different, this would suggest that the environment-based partitioning does not reflect functional differences. If the distributions do differ significantly, it would imply that the environmental features are related to the specific aspect of metabolism. Further, we computed the two-sample t-test for each individual map, module, operon, and COG. Those pathways that were significantly different (Benajamini-Hochberg corrected p < 0.05) were combined to form the DPM footprint.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.