In a CM, both the consensus structure of the model and a particular structural alignment of the model to an individual RNA sequence are binary trees.
For a pair-wise alignment against a consensus structure, output is as above except no consensus structure is produced.
80,000 models were generated and the lowest 5% (in full atom energy) were screened against the consensus structure.
For Rio2, the lowest 5% of models were used for screening against the consensus structure as described in the Results section.
The consensus structure shown in Figure 3 represents the typical position of specific and conserved amino acids or amino-acid categories.
As few as five geometric residue constraints derived from a consensus structure and further screening can select a highly accurate model.
The result is not only an alignment but a consensus structure that depicts residue conservation as a distribution of amino acids and amino-acid categories.
Comparisons between a sample and a consensus structure are very rapid (<1 second due to the reduced set of residues in a consensus set).
A secondary structure for each input sequence is predicted via RNAfold and a consensus structure is obtained by aligning the sequences using alifold [11].
The best model for E154 as determined against the consensus structure is shown in Figure 9A alongside the actual active-conformation structure (PDB code: 1ZAO).
We used fourteen active-conformation kinases (omitting Rio2) and generated a 52 residue consensus structure (consensus residues are listed in Supplementary Table S1).
