To overcome the dependence on the initial alignment by repeatedly
performing the MSA, estimating the value of some heruristic and
correcting the previous alignment until convergence is reached.
Global alignment:
- Simple iterative method: MulAlin, PRRP, DIALIGN programs
- Genetic algorithms (population of MSAs, mutation and
recombination events, selection)
- Hidden Markov Models
Local alignment (pattern discovery):
- Expectation-maximization
- Gibbs sampling
|