In phylogenetics, homology means..

Structural alignment is another perspective

Sequence alignment for phylogenetics is hard!

Sometimes it’s more or less impossible

Phylogenetic approach is best

Clustal W vs PRANK

PRANK is a smart “hack”

BAli-Phy is fully model-based

BAli-Phy is fully model-based

BAli-Phy is fully model-based

BAli-Phy is fully model-based

BAli-Phy is fully model-based

Note how BAli-Phy can trace the history of each site through
the tree, and the columns are simply a reflection of that.

BAli-Phy estimates alignment uncertainty

BAli-Phy is fully Bayesian, so it has a rigorous notion of alignment uncertainty, which can be elegantly summarized by that software:

BAli-Phy summary

• Co-estimate tree & alignment using Bayesian methods
• Can also use codon models for selection analysis; integrating out alignments eliminates artifacts
• Very computationally expensive: 60 sequences may take days

What’s funny about this protein-coding alignment?

What’s funny about this protein-coding alignment?

Multiple sequence alignment summary

• If you really care about alignments (and trees) use BAli-Phy
• MAFFT ranked the best in a recent benchmark by the authors of PRANK concerning ancestral sequence reconstruction
• If you want good alignment using PRANK
  with a lovely web-based GUI, use Wasabi
• If you want a quick and dirty alignment, use Muscle
  (esp. if your sequences aren’t too divergent)
• Do not use any variant of Clustal, even if your friends say to.
• Use a codon alignment if you have protein-coding genes.