Typical multiple sequence alignment tools, such as CLUSTAL W, MUSCLE, MAFFT, and T-COFFEE, do not handle indels accurately. The authors developed new tools, PRANK and PRANK_+F, that take into account the computed phylogeny of the sequences when placing insertions and deletions into the multiple alignment.

In this paper, the authors describe their refinements of the multiple alignment algorithm, and they provide theory and results that demonstrate that their algorithm improves the quality of multiple sequence alignments in a biologically meaningful way. The implications of their results are strongest for nucleotide sequence alignments, but the authors contend that their results are important also for peptide sequence alignments.

Source:

Löytnoja A, Goldman N. 2008. Phylogeny-aware gap placement prevents errors in sequence alignment and evolutionary analysis. Science 320:1632-1635. DOI: 10.1126/science.1158395.

Other bloggers who have have already commented on this paper include:

• anthropology.net
• Sandwalk

The Goldman group maintains a web page listing its publications; there are many other interesting papers given there.

Nimravid posted a summary, which was graciously acknowledged today by Dr. Gregory, and that was how I learned about the paper. Nimravid’s summary is good (as are all of Nimravid’s posts), but I recommend reading the entire paper, which is very readable and well-presented. I found that I have a pretty clear idea of how to interpret evolutionary trees, but I learned the differences in the terms dendrogram, cladogram, and phylogram.

Dr. Gregory makes an important point in his paper:

[I]t is impossible to know with certainty that any given phylogeny is historically accurate. As a result, any reconstructed phylogenetic tree is a hypothesis about relationships and patterns of branching and thus is subject to further testing and revision with the analysis of additional data.

In my mind, this is why evolution is a science. An evolutionary tree is a model of the evolutionary relationships among organisms. A good model is consistent with existing observations, and it provides a hypothesis than can be tested as additional data are gathered. If the model is inconsistent with the new data, the model is revised and subjected to testing with even more data.