We then aligned the sequences in each cluster using the local alignment algorithm implemented in DIALIGN [ 29], followed by the refinement algorithm from MUSCLE [ 30].
However, with the history reconstructed by FastML under random orthology assignments, the refinement algorithm only improves slightly over the base algorithm.
The refinement algorithm is given below.
Significant improvement was observed when applying the refinement algorithm once as compared with no refinement.
As shown in Section 3, the improved base-pairing probability matrices by the iterative refinement algorithm made the prediction accuracy much elevated, especially for short sequences.
The probability distribution produced by the iterative refinement algorithm can be regarded as a mixture of the probability distribution for each level p of the preceding prediction of pseudoknot-free structures, which can represent a wider space of the distribution than the individual distribution of pseudoknot-free structures.
The inferred networks were then refined by refinement algorithms with different models of gene duplications and losses.
Given the size of the tree and the root network, these are high rates of gene duplication and loss, yet, as we can see from Fig. 2, the improvement gained by our refinement algorithms remains clear in both plots, while RefineML further dominates RefineFast in both sensitivity and specificity, thanks to the appropriate reuse of the inferred leaf networks.
These two sources of error were acknowledged by Tsur et al. (2005), who briefly described a heuristic approach to account for 'shadows' (modifications that are misplaced by a PTM search engine), and later by the same group in the PTM refinement algorithm PTMFinder (Tanner et al., 2008).
In addition, it decreases the communication overhead and improves data locality by making use of a data partitioning scheme offered by the Parallel Delaunay Refinement algorithm (PDR).
