EVolutionary Couplings

The EVolutionary Couplings server provides functional and structural information about proteins derived from the evolutionary sequence record using methods from statistical physics.

  • Inputs alignments can be 'out the box' from PFAM for example, generated by this server using HHBlits or jackhmmer, or your own.
  • Uses maximum entropy algorithm, a global model, to calculate evolutionarily coupled residues to avoid transitive correlations.
    • Allows users to choose between a mean-field solution (EVcouplings-mfDCA)
    • OR The slower, but more accurate, pseudo likelihood maximization solution (EVcouplings-PLM) to the maximum entropy optimization.
  • Calculates which residues are the most evolutionary constrained in your protein.
    • In the graph of the network of residues connected by high ranking ECs, the scores on the edges are summed for every node.
  • Inputs the high ranking ECs as close distance restraints into a standard NMR module in the CNS_solve environment.
    • This module applies distance geometry using the restraints, on an extended polypeptide, followed by short runs of simulated annealing to produce a set of candidate models.(Usually about 200 models are sufficient)
  • Ranking of models is based on simple geometric criteria and clustering.

What is EVcouplings?

EVcouplings calculates ECs for a protein of interest if there are enough sequences in the corresponding protein family. The ECs reflect a network of interactions between residue pairs based on the evolutionary information in the family. The most strongly constrained interactions provide insight into mechanisms of protein function. You can view the network of interactions between residue pairs without having a 3D structure. If a 3D structure is available, you can also view the network of residue-residue interactions as lines connecting the residue pairs. The constraint strength for single residues is informative beyond single 'column' conservation in a multiple sequence alignment. If a 3D structure is available you can view the constraint strength as a single-residue property, mapped in 3D either as spheres or in a sausage/putty representation.

  • Enter a protein sequence and a known PDB ID
  • Optionally, provide your own multiple sequence alignment. For alpha-helical transmembrane proteins you can also provide your own topology prediction.
  • Get EC scores for residue pairs, EC strength for individual residues and contact maps of high-ranking pairs.
  • Get evolutionary couplings painted on the 3D structure
  • View ECs painted on the known 3D structure

Start >>

What is EVfold?

EVfold uses evolutionary variation to calculate a set of co-evolved residue pairs in a protein family using a global approach called maximum entropy, formally similar to partial correlations. The ECs are calculated rapidly using EVcouplings-mfDCA (mean-field direct coupling analysis) or more slowly and accurately with EVcouplings-PLM. The resulting residue pairs represent a network of interactions across the protein and reflect co-evolution at pairs of positions during the evolutionary trajectory of the protein. The ECs are used as distance constraints on an unfolded protein, which is then folded using distance geometry and simulated annealing. EVfold handles both globular and transmembrane proteins and is currently optimized for individual domains.

  • Enter a protein sequence.
  • Optionally, provide your own multiple sequence alignment. For alpha-helical transmembrane proteins you can also provide your own topology prediction.
  • Get EC scores for residue pairs, the EC strength for individual residues and contact maps of high-ranking pairs.
  • Explore a set of ranked predicted 3D structure models

Start >>

What is EVcomplex?

EVcomplex is a method for predicting inter-protein residue interactons. Two multiple sequence alignments for two protein domains are combined by concatenating sequence pairs, one from each of the domain alignments. Pairing is done by genome coordinates; two sequences in the same genome from nearby locations are most likely to be co-expressed and physically interact. These concatenated pairs are then used as an MSA input into the Evolutionary Couplings method, scoring pairs of residues based on evolutionary co-variation. High scoring ECs are examined to distinguish the inter-ECs (which pair two residues from different domains) from the intra-ECs. The high scoring inter-ECs are likely points of interaction in the interface between the two domains.

  • Results and discussions from recent publication are available.
  • Interactive server (in beta test) is available where you can:
    • Enter the sequences of two domains which are known to interact, and have known bacterial homologs.
    • Provide other optional inputs similar to those for EVCoupling
    • Get the concatenated pair multiple sequence alignment
    • Get the high scoring inter-ECs
    • Visualize inter-ECs on known 3D structures of the protein complex.

Start >>

The evolutionary couplings approach was developed collaboratively between
the Marks lab in the Department of Systems Biology at Harvard Medical School(HMS) and
the Sander lab in the Computational Biology Center (cBio) at Memorial Sloan-Kettering Cancer Center(MSKCC).