Adaptive Local Realignment of Protein Sequences

Dan Deblasio, John Kececioglu

Research output: Contribution to journalArticlepeer-review


While mutation rates can vary markedly over the residues of a protein, multiple sequence alignment tools typically use the same values for their scoring-function parameters across a protein's entire length. We present a new approach, called adaptive local realignment, that in contrast automatically adapts to the diversity of mutation rates along protein sequences. This builds upon a recent technique known as parameter advising, which finds global parameter settings for an aligner, to now adaptively find local settings. Our approach in essence identifies local regions with low estimated accuracy, constructs a set of candidate realignments using a carefully-chosen collection of parameter settings, and replaces the region if a realignment has higher estimated accuracy. This new method of local parameter advising, when combined with prior methods for global advising, boosts alignment accuracy as much as 26% over the best default setting on hard-to-align protein benchmarks, and by 6.4% over global advising alone. Adaptive local realignment has been implemented within the Opal aligner using the Facet accuracy estimator.

Original languageEnglish (US)
Pages (from-to)780-793
Number of pages14
JournalJournal of Computational Biology
Issue number7
StatePublished - Jul 2018


  • alignment accuracy
  • iterative refinement
  • local mutation rates
  • multiple sequence alignment
  • parameter advising

ASJC Scopus subject areas

  • Modeling and Simulation
  • Molecular Biology
  • Genetics
  • Computational Mathematics
  • Computational Theory and Mathematics


Dive into the research topics of 'Adaptive Local Realignment of Protein Sequences'. Together they form a unique fingerprint.

Cite this