Accurate reconstruction of the branching order for the major eukaryotic lineages is an extremely challenging task given the low information content of alignments of highly diverged sequences and the compressed cladogenesis that seems to be characteristic of the primary radiation of eukaryotes [ 5- 7].
Many times there is no need to modify the contents of alignment blocks, but instead it is desirable to apply a set of filters that remove blocks that do not match a set of conditions.
These metrics include yield, alignment and duplication rates; GC bias, rRNA content, regions of alignment (exon, intron and intragenic), continuity of coverage, 3′/5′ bias and count of detectable transcripts, among others.
A good synthetic dataset should reproduce empirically-observed features of actual alignments, including gap (indel) structures and local correlations between nucleotides, both of which locally deplete the information content of an alignment and can elevate false-positive rates.
These results were then filtered by information content of the alignment in the same way as the SWISS-PROT mutagenesis analysis to result in a list of ∼22,000 SNPs.
Ref. [ 32] suggested another way to estimate the phylogenetic information content of an alignment.
However, it has never been asked how to test if the information content of an alignment is sufficient to find a realistic model.
Thus, circular orders appear to offer a robust way to assess the "phylogenetic information content" of characters (alignment columns) without strongly prejudicing the subsequent tree construction.
These hypotheses can refer (a) to the monophyly of a group of species or (b) to the phylogenetic information content of an alignment (e.g. "gene A is better than gene B to infer the history of taxon X").
Note that in the last cycle of search, the algorithm adds/subtracts each segment from query sequences based on its impact to the informational content of the alignment [ 15].
