This page implements the contact prediction method described in:
Using inferred residue contacts to distinguish between correct and incorrect protein models. Christopher S. Miller and David Eisenberg Bioinformatics 2008 (Advance Access - doi: 10.1093/bioinformatics/btn248)
[PubMed]
[Bioinformatics OpenAccess]
Notes about multiple sequence alignments
The sequence for which you would like contact predictions should be the first sequence in your multiple sequence alignment.
The contact prediction method was tested on multiple sequence alignments built by gathering BLAST homologs with E-value <= 1e-10. We suggest you follow a similar protocol.
Your alignment will be pre-filtered to remove sequences with > 80% identity to another sequence in the alignment.
Download the source code of the software to run locally here