Structure Alignments

This Java implementation for multiple structure alignments, named MultipleMC, is based on the original CE-MC implementation by [Guda C, Scheeff ED, Bourne PE & Shindyalov IN in 2001](http://psb.stanford.edu/psb-online/proceedings/psb01/abstracts/p275.html)

[](http://www.ncbi.nlm.nih.gov/pubmed/11262947).

The idea remains unchanged: perform **all-to-all pairwise alignments** of the structures, choose the

**reference** as the most similar structure to all others and run a **Monte Carlo optimization** of

the multiple residue equivalencies (EQRs) to minimize a score function that depends on the inter-residue

distances.

Although the main idea is the same as in the original algorithm, some details of the implementation have

been changed in the BioJava version. They are described in the main class, but as a summary:

1. It accepts **any pairwise alignment** algorithm (instead of being attached to CE), so any

of the algorithms described before is suitable for generating a seed for optimization. Note that

this property allows *non-topological* and *flexible* multiple structure alignments, always restricted

by the pairwise alignment algorithm limitations.

2. The **moves** in the Monte Carlo optimization have been simplified to 3, instead of 4.

3. A **new move** to insert and delete individual gaps has been added.

4. The scoring function has been modified to a **continuous** function, maintaining the properties that the authors described.

5. The **probability function** is normalized in synchronization with the optimization progression, to improce the convergence into a score maximum after some random exploration of the multidimensiona space.

The algorithm performs similarly to other multiple structure alignment algorithms for most protein families.

The parameters both for the pairwise aligner and the MC optimization can have an impact on the final result. There is not a unique set of parameters, because they usually depend on the specific case. Thus, trying some parameter combinations, keeping in mind the effect they produce in the score function, is a good practice when doing structure alignments.

BioJava class: [org.biojava.nbio.structure.align.multiple.mc.MultipleMcMain]

(www.biojava.org/docs/api/org/biojava/nbio/structure/align/multiple/mc/MultipleMcMain.html)

The **pairwise structure alignment** algorithms in BioJava implement the

`StructureAlignment` interface, and are usually accessed through

StructureAlignmentDisplay.display(afpChain, ca1, ca2);

For creating **multiple structure alignments**, the code is a little bit different, because the

returned data structure and the number of input structures are different. Here is an

example of how to create and display a multiple alignment:

List<String> names = Arrays.asList("3app", "4ape", "5pep", "1psn", "4cms", "1bbs.A", "1smr.A");

AtomCache cache = new AtomCache();

List<Atom[]> atomArrays = new ArrayList<Atom[]>();

for (String name:names) {

atomArrays.add(cache.getAtoms(name));

}

StructureAlignment pairwise = StructureAlignmentFactory.getAlgorithm(CeMain.algorithmName);

MultipleMcMain multiple = new MultipleMcMain(pairwise);

MultipleAlignment result = algorithm.align(atomArrays);

System.out.println(MultipleAlignmentWriter.toFASTA(result));

MultipleAlignmentDisplay.display(result);

| [Book 3: The Structure modules](README.md)