Browse by
Summary Table
Presenting Author
All Authors

Abstract Detail

Methods and theory of phylogenetic inference

Kelchner, Scot A. [1], Swofford, David L. [2], Wilgenbusch, James [2].

Evaluating partitioned models for phylogenetic analysis of combined data..

The use of combined data sets for phylogenetic inference is now customary in molecular systematics. This is largely due to an expectation that more data will provide improved resolution and accuracy. However, the typical strategy of applying a uniform model of character evolution to a combined data matrix could potentially decrease the accuracy of phylogeny estimation. This is because combined data often represent multiple process partitions that evolve under different mutational constraints (e.g., protein-coding DNA, rDNA, morphology, indels). Applying the wrong model of character evolution to such data can result in systematic error, which in turn may increase the chance of selecting a wrong tree. A question of interest is whether a model (such as parsimony or GTR) that is applied uniformly to a combined data set will perform worse due to systematic error than a partitioned model which better accommodates dissimilar mutation patterns among data partitions. We investigated this question with a computational experiment in which combined molecular data sets were simulated on various tree structures using different models of evolution for each process partition. The heterogeneous data sets were then analyzed with uniform models and with partitioned likelihood models that estimate separate parameter values for each character partition. We compare the accuracy of the two approaches for different tree structures and quantities of data. The results have important implications for classic problems in phylogenetic analysis including long branch attraction, inferred incongruence between partitions, and outgroup rooting.

1 - Australian National University, School of Botany and Zoology, Bldg 116, Canberra, Australian Capital Territory, 00200, Australia
2 - Florida State University, Computational Science & Information Technology, Dirac Science Library, Tallahassee, Florida, 32306, USA

phylogenetic inference
systematic error
outgroup rooting
partitioned models
combined data.

Presentation Type: Symposium
Session: 39-2
Location: Alpine A (Snowbird Center)
Date: Tuesday, August 3rd, 2004
Time: 3:00 PM
Abstract ID:132

Copyright © 2000-2004, Botanical Society of America. All rights reserved.