A comparison of supermatrix and supertree methods for multilocus phylogenetics using organismal datasets

Area of Science:

• Phylogenetics
• Computational Biology
• Bioinformatics

Background:

• Supertree methods are proposed for computational tractability in large multilocus phylogenomic datasets.
• These methods involve constructing locus-specific trees and combining them.
• Previous studies using simulated data suggested supertrees are efficient and accurate.

Purpose of the Study:

• To compare the accuracy (tree length) and computational time of supertree methods versus supermatrix approaches.
• To evaluate these methods using real organismal data, not just simulations.

Main Methods:

• Compared supertree (SuperFine, SuperTriplets) and supermatrix (heuristic search in TNT) methods.
• Utilized 20 multilocus datasets from organismal data.
• Assessed tree length using the parsimony criterion and compared computational times.
• Statistical analysis employed the Wilcoxon matched-pairs signed rank test.

Main Results:

• Supermatrix searches resulted in significantly shorter trees compared to both supertree methods (P < 0.0002).
• Supermatrix search was significantly faster than SuperFine + locus-specific search (P < 0.01).
• Supermatrix search computational time was equivalent to SuperTriplets + locus-specific search (P > 0.4).

Conclusions:

• For real phylogenomic data, supertree methods offer no advantage over supermatrix approaches in terms of time or tree length.
• Heuristic search using a supermatrix is recommended over supertree methods for large phylogenomic analyses.