increasing human grasp of GOD...,

wired | At the most basic level, scientists create phylogenetic trees by grouping species according to their degree of relatedness. Lining up the DNA of humans, chimpanzees and fish, for example, makes it readily apparent that humans and chimps are more closely related to each other than they are to fish.

Researchers once used just one gene or a handful to compare organisms. But the last decade has seen an explosion in phylogenetic data, rapidly inflating the data pool for generating these trees. These analyses filled in some of the sparse spots on the tree of life, but considerable disagreement still remains.

For example, it’s not clear whether snails are most closely related to clams and other bivalves or to another mollusk group known as tusk shells, said Rokas. And we have no idea how some of the earliest animals to branch off the tree, such as jellyfish and sponges, are related to each other. Scientists can rattle off examples of conflicting trees published in the same scientific journal within weeks, or even in the same issue.

“That poses a question: Why do you have this lack of agreement?” said Rokas.

Rokas and his graduate student Leonidas Salichos explored that question by evaluating each gene independently and using only the most useful genes — those that carry the greatest amount of information with respect to evolutionary history — to construct their tree.

They started with 23 species of yeast, focusing on 1,070 genes. They first created a phylogenetic tree using the standard method, called concatenation. This involves stringing together all the sequence data from individual species into one mega-gene and then comparing that long sequence among the different species and creating a tree that best explains the differences.

The resulting tree was accurate according to standard statistical analysis. But given that similar methods have produced trees of life that are rife with contradiction, Rokas and Salichos decided to delve deeper. They built a series of phylogenetic trees using data from individual yeast genes and employed an algorithm derived from information theory to find the areas of greatest agreement among the trees. The result, published in Nature in May, was unexpected. Every gene they studied appeared to tell a slightly different story of evolution.

“Just about all the trees from individual genes were in conflict with the tree based on a concatenated data set,” says Hilu. “It’s a bit shocking.”

They concluded that if a number of genes support a specific architecture, it is probably accurate. But if different sets of genes support two different architectures equally, it is much less likely that either structure is accurate. Rokas and Salichos used a statistical method called bootstrap analysis to select the most informative genes.

In essence, “if you take just the strongly supported genes, then you recover the correct tree,” said Donoghue. Fist tap Dale.