Evolutionary biologists are often interested in reconstructing how different genes evolved from each other. Large numbers of genes can now be sequenced quickly but the development of statistical methods has lagged behind. To analyse even moderately large data sets under realistic evolutionary models, researchers have been forced to use supercomputers.
But this is now changing quickly since evolutionary biologists discovered an old statistical approach called Bayesian inference.
Reverend Thomas Bayes formulated the basic principles already in the 18th century but the approach was not feasible for most problems until computers could be used to solve the difficult equations numerically.
Scientists from the US and from the Evolutionary biology Centre at Uppsala University have now developed computer-based methods for Bayesian inference of problems in evolutionary biology.
The basic idea is to allow the computer to walk around, according to certain rules, in a landscape defined by the relative probability of different evolutionary scenarios. By allowing the computer to jump between a normal landscape and heated landscapes, in which hilltops and valleys are less clearly separated, the computer can quickly find the hilltops, corresponding to the most likely evolutionary scenarios.
In a paper published in Science, December 14, the Swedish-American group shows how problems that have previously been too difficult for supercomputers can now be analysed using an ordinary desktop computer. The new methods can be applied to a wide range of problems, including reconstruction of relationships among species, clarification of viral transmission pathways, and studies of the mechanisms of molecular evolution.
The above post is reprinted from materials provided by Uppsala University. Note: Content may be edited for style and length.
Cite This Page: