How to run partiFold?

Example output webpage

How to choose structural constraints?

The related partiFold papers (see references) give a more thorough description of the execution parameters, however, this quick summary should be enough to get up and running. To begin, transmembrane β-barrel structures can be constrained by a number of biologically motivated parameters. These structural constraints aims to fit these features by defining the allowable conformation space. These include:

  1. number of TM β-strands in the barrel,
  2. length of TM β-strands,
  3. strand inclination with respect to membrane plane (shear number),
  4. size of periplasmic and extra-cellular loops,
  5. hydrophobic profile of TM β-strands.

How to choose scoring parameters?

In addition, the scoring function section allows the user to customize the pseudo-energy model used for the scoring conformations. Here, a menu option allow one to choose specific sets of parameters for:

  1. Stacking pair contact potentials (classical pairwise potential can also be used),
  2. polarity scale used to (optionally) filter loops between strand pairs,
  3. hydrophobic scale used for selecting TM β-strands,
  4. physical (Boltzmann) constants.

Job control

Finally, the job controls section allow you to (i) restrict your prediction to our previous method transFold by computing ONLY the minimum folding energy structure (Note: this is a subset of the predictions achieved now by partiFold), (ii) choose the number of sample TMB structures and (iii) provide your email address if you want to be receive your prediction through email.

How to read partiFold output?

Multiple outputs are available for these predictions. We describe these below. You may want first to have a look to an example output webpage to familiarize yourself.

partiFold (Single structure) standard output:

This output displays the minimum folding energy super-secondary structure as defined in transFold (See references). The same notation is used to display sampled structures.

The prediction is displayed in three lines. The first line contains the amino acid sequence input by the user. The second, third and fourth lines contain structure predictions made by partiFold: secondary structure, topology and inter-residue contacts. Obviously, each of these lines describes the same secondary structure! Residues denoted as E, C or M are predicted to belong to a transmembrane β-strand (C denotes "channel", i.e. facing the cavity of the β-barrel protein, while M denotes "membrane", i.e. facing the outer membrane bilayer). All other positions are predicted to be loop positions -- either in the extra-cellular loop or the intra-cellular loop region.


The topology prediction (i.e. the orientation of the transmembrane strands through the membrane) is given by the notation used for the amino acids located in turn regions. Residues which are predicted to be inside cell (or in the periplasm milieu) are marked with i, and those exposed to the extra-cellular environment are denoted with o. The label E is used to mark strand extensions.

Residue-contacts are denoted by paired residues of the same type (C or M) between second and third lines. Pairings are articulated around a turn (denoted i or o). The first residue on the left of the third line is paired the first residue on the right of the second line. The second residue on the left of the third line is paired with the second residue on the right of the second line. and soon. Notation is inverted for the closing pair. The figure below illustrates this situation.

2-tape output example

Stochastic contact map:

The figure provides a graphical representation of the inter-strand per-residue contact probabilities. Rows and columns are annotated with the sequence. The entry (i,j) gives the probability P(i,j) of a contact between the i-th and j-th residues. The surface of the black dots associated with these entries illustrates the magnitude of the probability. Thus a blank entry represents a probability of zero.

The entries above the diagonal are for interactions occurring between amino-acids with a side chain pointing outward the channel. While those below the diagonal are for contacts occurring inside the pore.

Per-residue contact probability and entropy:

These curves display the propensity of a residue to be involved in an inter-strand β-sheet interaction. These statistics are computed from the residue contact probabilities. The per-residue contact probability P(i) of a residue at index i is computed as ∑j P(i,j). Note that since a residue may be involved in two distinct interactions (each strand is involved in two anti-parallel pairings) this value belongs to [0,2]. Similarly, the per-residue contact entropy E(i) of a residue at index i is computed as ∑j -log(P(i,j)).

These values are displayed in a graph where the x-axis is the sequence index and the y-axis the per-residue contact probability or entropy.

Sampled structures from Boltzmann ensemble sampling:

The sampled TMB structures are displayed using the 2-tape representation described above. For each sampled structure, its probability in the Boltzmann ensemble is also provided.

> TMB sample 0 (probability in Boltzmann ensemble: 3.601682e-06) .........EEEMCMCMCMoooMCMCMCMCMCiiiiCMCMCMCMCMEEEEoooooooMCMCMCMCMCMCMCEEiiCMCMCMCMCMCoooooooooooo ... .........CMCMCMCMCMoooMCMCMCMCMCiiiiCMCMCMCMCMCMCMoooooooEEEEECMCMCMCMCMCiiCMCMCMCMCMCoooooooooooo ... > TMB sample 1 (probability in Boltzmann ensemble: 3.593129e-08) ..EEMCMCMCMEoooooooooCMCMCMCMCMEEiiMCMCMCMCMCoooooooooooooooooooCMCMCMCMCMEiiiiCMCMCMCMCMCEEEooooo ... ..MCMCMCMCMCoooooooooEECMCMCMCMCMiiMCMCMCMCMCoooooooooooooooooooCMCMCMCMCMCiiiiCMCMCMCMCMCMCMooooo ...

partiFold downloadable output files:

All predictions detailed above can be downloaded. First a summary of the folding constraints and folding parameters used for the prediction is provided in a text file. The sampled TMB structures are also proposed in a similar format.

The stochastic contact map is provided as a pdf document. The individual contact probabilities values are also available in a text format. The first row contains the number of residue in the sequence. The probabilities are given a 4-column format where the two first columns gives the index of the residue, the third one the environment of the interaction (channel C or membrane M) and the fourth one the numerical value.

The graphs of the per-residue contact probabilities and entropy are also provided in pdf format. Numerical values can be downloaded in a 2-column text format, where the first column give the residue index and the second the per-residue value.

The single minimum folding energy super-secondary structure (classical output) is provided in text format. Additionally, a file summarizing the prediction in a more traditional format can be dowloaded as a 5-column, tab delimitated, text file. Each row of this file corresponds to a residue. The first column contains the index of the current amino acid; the second column contains the single letter amino acid code associated with this residue (all non-IUPAC single letter amino acid codes are stripped). The third column contains the secondary structure residue assignment as well as the side-chain orientation of TM β-strand residues. Hence, a residue marked as M or C is predicted to belong to a TM β-strand, where C denotes “channel” (i.e. facing the cavity of the β-barrel protein) and M denotes “membrane” (i.e. facing the outer membrane bilayer). A residue marked as i is predicted to be in the periplasm, a residue marked as o to be extra-cellular, and a residue marked as “.” to be exterior to the membrane, but not in a turn. The last two columns are only used for TM β-strand residues and give the index of the amino acids interacting with the current one (including the interaction of the closing strand pairing).

Evaluation of partiFold on known PDB structures