Online activity was logged.
All deletions are logged.
As Last is based on this generalized scoring model, the score assigned by Last to an alignment is the log of its likelihood ratio.
Given a substitution model, there is an alignment scoring scheme that corresponds to that substitution model, so that the score for any alignment is the log of the likelihood of the corresponding set of substitutions.
To compare the UD-trees with the "best tree", the alignment was loaded into PAUP and site-wise log-likelihood values for each tree were calculated.
In the score reconstruction/dissection/statistical re-evaluation/hits classification stage, the full score of each alignment is reconstructed using the emission/transition/invariant log-odd scores of the alignment's domain model.
Finally, we decide if the alignment is more likely to represent a protein-coding region or a non-coding region based on the log-likelihood ratio.
"The alignment is great".
Alignment is a tricky task.
Obviously, alignment is critical.
The alignment is deliberate.
