To properly understand protein function, one should consider evolution, sequence alignment, and structural similarities. We've advanced the science of evolutionary (phylogenetic) tree construction, by considering more realistic models. One new phylogenetic tree model allows nonconvex characters with an associated penalty. Another models polymorphism (multiple states per character per species). We have also defined a new form of consensus tree (combining multiple trees on the same data) that captures more information than current popular models. Our new multiple sequence alignment cost function encourages full and fair interaction among all sequences. Alignments scoring well under this metric may reveal fundamentally different relationships among protein sequences than alignments that score well under metrics where some small set of sequences can dominate.