After finding CALs for each read, we use a modified local alignment algorithm previously described for color space reads by Homer et al.[17] and Rumble et al. [23].
For long reads, a possibly better solution would be to divide the read into multiple short fragments, align the fragments separately with the algorithm described above and then join the partial alignments to get the full alignment of the read.
The truncated sequence local alignment algorithm we describe here, for finding an optimal local parse tree π gh that accounts for a linear sequence fragment, does require a substantial modification of the CYK algorithm because it needs to identify the optimal endpoints g, h.
We have implemented the algorithm described above in the C++ language as the Alignment with Gap Excision (AGE ) program (freely available at http://sv.gersteinlab.org/age).
Then, a dendrogram is built, using the parallel neighbor-joining clustering algorithm described earlier, from all the pair-wise DP alignment scores from step (i).
We then apply the minimum normalized energy folding algorithm described in the beginning of this section to the refined alignment A, treating each pairing column in A as a possible base pair considered in the folding algorithm.
Alignment of peptide features between LC-MS/MS runs was performed within ProSE using the built-in algorithm described by Sandin, et al. [ 74].
The alignment then proceeds by sequentially registering the remaining subjects to the group atlas using the pair-wise functional registration algorithm described above.
To improve the Needleman-Wunsch and Gotoh alignments at the ends of sequences that do not fully overlap, I followed the end-space free variant algorithm described by Gusfield [34].
The alignments for the 70 sequences in the related data set are extracted from the STAP database, and the query sequence is aligned to them using the CLUSTALW profile alignment algorithm [40] as described above for domain assignment.
