ENCoM is described, validated and applied in various contexts in the following publications:
Frappier, V., & Najmanovich, R. J. (2014). A coarse-grained elastic network atom contact model and its use in the simulation of protein dynamics and the prediction of the effect of mutations.
PLoS Computational Biology, 10(4)PubMed
| Plos Computational Biology
| Cited by
ENCoM was applied to the study of thermophile and mesophile proteins:
Frappier V, Najmanovich RJ (2014) Vibrational entropy differences between mesophile and thermophile proteins and their use in protein engineering
. Protein Science. doi:10.1002/pro.2592.PubMed
| Protein Science
The ENCoM server aims to give access to applications of ENCoM for non-technicl users. It presents two interfaces for 2 applications:1. The prediction of the effect of mutations
, to evaluate the effect of thermostability for multiple mutations at a time2. The conformational sampling
, that allows the generation of geometrically realistic conformations.
In both interfaces case, there are 4 steps
- Input Structure: Upload a file (PDB format) or enter a 4-letter PDB code to define the structure to use as query. Every structure analysed with ENCoM is stripped of heteroatoms and hydrogen atoms.
- Chains:: ENCoM takes in consideration inter-chain contacts for all selected chains to calculate modes. For chain-specific contacts only, you can specify here the list (comma separated) of the chains to consider. Leave blank for all chains.
- Mutations / Parameters: for the mutations interface, define which mutations to evaluate; for conformation sampling, determine parameters such as number of mode and maximum distortion. See the respective sections for details and examples.
- Email Address (optional): a notification will be sent to this adress upon to job submission and termination.
Please note that when using PDB accession codes, it is important to note that the ENCoM server uses the asymmetric unit, thus it is important to specify the correct chains to be included in the calculations or alternatively download the biological unit independently from a PDB repository and use this as input.
Upon the submission of a job the user receives an email with a link to a result page where the results (and related files) will remain for 7 days. There, a log shows the level of completion of the job. Run time varies from minutes to several hours depending on the complexity of the job and server load.
To specify which chain-specific positions and the mutations to evalute at these positions three elements are needed:
- residue position or range (25 or 1-10)
- chain of the specified positions (single letter e.g. A)
- amino-acids mutations to evalute (one letter code, comma separated or "all"
Below are some examples of input strings that can be entered at step 3.25 A W
A simple mutation of residue number 25 in chain A to a tryptophan.50-100 A W
Similar to the first example but mutating individually positions 50 to 100 inclusively.25 A W,H,A
Mutating residue of position 25 in chain A to tryptophan, histidine and alanine.25 A all
Mutating residue 25 in chain A to the remaining 19 amino-acids.1-1000 A A
This performs an alanine scan of the chain A of the structure (ENCoM skips residues numbers between 1-1000 that do not exist)
You can specify multiple mutations for a query by separating the mutations with a semi-colon (;).25 A W; 50-100 A H,W
Mutating residue 25 in chain A to the remaining 19 amino-acids.1-1000 A A; 1-100 B A; 1-100 C A
This performs an alanine scan of chains A, B and C.Results Table
The results table depicted below is given for each job. It shows all evaluated mutations, the predictions by vibrational entropy based ENCoM, the enthalpy-based prediction of FOLDX and their combined score (all in ΔkCal/mol). A GIF file shows a tilting 3D structure in cartoon, colored based on the effect of the mutation on the structure (red more flexible, blue more rigid). A static image (PNG) and a PyMOL session for this colored structure is also given along with the mutant PDB structure.Downloadable Results
The mutant model structures, pymol sessions and result graphs are individually downloadable:sessions.zip
folder contains a PyMOL script for each mutant to colour the modelled structure according to the calculated ΔSvib
values: from blue to red representing more rigid to more flexible with respect to the wild type.
directory contains all the coordinate files for each modelled mutant.seq_tolerance.pdf
shows the calculated ΔSvib
values for mutation in each position clustered by the residue mutated into. This is particularly interesting when performing all mutations in all positions.top25_graph.pdf
same as all_graphs.pdf
but as the name suggests, only for the top 25 mutations.all_ddg.cvs
contains the predicted effect of mutations. In this file you will find one mutation per row. For each mutation we show the scaled ΔΔG calculated with ENCoM, FoldX and the combined value (their sum). We provide the overal rank of the mutation (i.e., from 1 to N, where N is the number of mutations) in terms of the overall rank (stabilizing to destabilizing) as well as the mutation's rank within that position (1 to at most 19). The second row of the file represents the "wild type" sequence used as input. Each collumn after the position rank represents one particular position in the sequence. For each row, including the wild type, the value presented shows the calculated vibrational entropy. Vibrational entropy differences (ΔSvib
) for any residue in any mutant are obtained from the subtraction of the respective values. We provide the data in this format to allow the calculation of vibrational entropy differences between mutants.The more negative this value the more rigid this position becomes with respect to reference state used.all_graphs.pdf
file contains a plot of ΔSvib
for each position in the sequence for each mutant with respect to the wild type. The distribution of ΔΔG for all mutants as well as a heatmap for the ΔΔG values and the scaled ΔΔG from ENCoM and FoldX are also provided as graphs. In the later, each row represents one mutant.
There are four parameters to set that will affect the generated structures:
- the number of modes used to sample the conformational space (e.g. 3)
- the starting mode (minimum: 7, see note below)
- the maximum structural distortion in Angstroms (e.g. 2)
- the RMSD step in Angstroms per conformation
Please note that normal modes (eigenvectors) are calculated using the chains of the input structure provided. These modes are ordered by their respective eigenvalue. The first 6 normal modes correspond to translation and rotations and are not relevant. Any conformation generated is the result of a linear combination of amplitudes assigned to eigenvectors (modes). Thus, the more modes chosen, the more realistic the generated conformations will be. The lower the rank of normal modes (minimum is 7), the more accessible this mode of movement is. However, it also represents more global movements. Thus in reality a smaller "amount" of movement in that direction is generated around the structure used to define the equilibrium around which the quadratic potential approximation used in the calculation of the normal modes is valid.Uniform Sampling
Whereas we present the movement of each normal mode individually in the mode_motion
directory as multi-state PDB files, these are unrealistic in that they never occur in isolation and not with the magnitudes shown. As we strive to generate a uniform sampling (in terms of RMSD) of the conformational space within the choice of normal modes selected by the user, the number of such conformations increases exponentially.
Therefore, depending on the maximum RMSD from the equilibrium structure and the minimum RMSD distortion between conformations that the user defines, we try to sample uniformely the combination of normal modes and may adjust some of the parameters chosen by the user to achieve a maximum of 250 conformations generated.Geometrically realistic structures
While normal modes analysis such as the present one generates movements in Cartesian space, the resulting conformations do not respect any constraints on bond angles and distances. We rebuilt each conformation using Modeller with the normal-mode generated conformation as a template, thus implicitly mapping the unrealistic structures generated into real ones.Presented Results
The screenshot below show the resuts for a given job. The image on the left is an animated GIF file aside to a PNG image the combined generated structures. The images are followed by a table that displays the scalar amplitude of each mode for each model.Description of the files in Results.zipall_models.pdb
all conformations ordered in a way to mimimize the RMSD between consecutive conformation to give a semblance of a path. It must be noted that this is for visualisation purposes only and may not represent in any way an actual dynamics trajectory of the protein.amplitudes.dat
The amplitude of the modes selected for every model.images
Contains a GIF and PNG of the combined models.mat.dat
A matrix of the pairwise RMSD between structures. Model 1 is the original model.Mode_motion folder
The motions for modes 7-20 in PDB format.Models folder
Each generated conformation in PDB format.
When all_models.pdb is opened, click the button indicated in the screenshot below to animate the models.
The ENCoM server was developped by Matthieu Chartier
, Vincent Frappier
and Rafael Najmanovich
. See the contact page
to drop us a message.