JABAWS Protein Disorder Prediction Services
The Web Services→Disorder menu in the alignment window allows access to protein disorder prediction services provided by the configured JABAWS servers. Each service operates on sequences in the alignment or currently selected region (since Jalview 2.8.0b1) to identify regions likely to be unstructured or flexible, or alternately, fold to form globular domains.

Predictor results include both sequence features and sequence associated alignment annotation rows. Features display is controlled from the Feature Settings dialog box. Clicking on the ID for a disorder prediction annotation row will highlight or select (if double clicked) the associated sequence for that row. You can also use the Sequence Associated option in the Colour By Annotation dialog box to colour sequences according to the results of predictors shown as annotation rows.

JABAWS 2.2 provides four disorder predictors which are described below:

DisEMBL (Linding et al., 2003)
DisEMBL is a set of machine-learning based predictors trained to recognise disorder-related annotation found on PDB structures.

Name Annotation type Description
COILS Sequence Feature &
Annotation Row
Predicts loops/coils according to DSSP definition[1].
Features mark range(s) of residues predicted as loops/coils, and annotation row gives raw value for each residue. Value over 0.516 indicates loop/coil.
HOTLOOPS Sequence Feature &
Annotation Row
"Hot loops constitute a refined subset of COILS, namely those loops with a high degree of mobility as determined from Cα temperature factors (B factors). It follows that highly dynamic loops should be considered protein disorder."
Features mark range(s) of residues predicted to be hot loops and annotation row gives raw value for each residue. Values over 0.6 indicates hot loop.
REMARK465 Sequence Feature &
Annotation Row
"Missing coordinates in X-ray structure as defined by remark465 entries in PDB. Nonassigned electron densities most often reflect intrinsic disorder, and have been used early on in disorder prediction."
Features gives range(s) of residues predicted as disordered, and annotation row gives raw value for each residue. Value over 0.1204 indicates disorder.

[1]. DSSP Classification: α-helix (H), 310-helix (G), β-strand (E) are ordered, and all other states (β-bridge (B), β-turn (T), bend (S), π-helix (I), and coil (C)) considered loops or coils.

RONN a.k.a. Regional Order Neural Network
This predictor employs an approach known as the 'bio-basis' method to predict regions of disorder in sequences based on their local similarity with a gold-standard set of disordered protein sequences. It yields a set of disorder prediction scores, which are shown as sequence annotation below the alignment.

Name Annotation type Description
JRonn[2] Annotation Row RONN score for each residue in the sequence. Scores above 0.5 identify regions of the protein likely to be disordered.

[2]. JRonn denotes the score for this server because JABAWS runs a Java port of RONN developed by Peter Troshin and distributed as part of Biojava 3

IUPred employs an empirical model to estimate likely regions of disorder. There are three different prediction types offered, each using different parameters optimized for slightly different applications. It provides raw scores based on two models for predicting regions of 'long disorder' and 'short disorder'. A third predictor identifies regions likely to form structured domains.

Name Annotation type Description
Long disorder Annotation Row Prediction of context-independent global disorder that encompasses at least 30 consecutive residues of predicted disorder. Employs a 100 residue window for calculation.
Values above 0.5 indicates the residue is intrinsically disordered.
Short disorder Annotation Row Predictor for short, (and probably) context-dependent, disordered regions, such as missing residues in the X-ray structure of an otherwise globular protein. Employs a 25 residue window for calculation, and includes adjustment parameter for chain termini which favors disorder prediction at the ends.
Values above 0.5 indicate short-range disorder.
Structured domains Sequence Feature Features highlighting likely globular domains useful for structure genomics investigation.
Post-analysis of disordered region profile to find continuous regions confidently predicted to be ordered. Neighbouring regions close to each other are merged, while regions shorter than the minimal domain size of at least 30 residues are ignored.

Defines regions of globularity or natively unstructured regions based on a running sum of the propensity of residues to be structured or unstructured. The propensity is calculated based on the probability of each amino acid being observed within well defined regions of secondary structure or within regions of random coil. The initial signal is smoothed with a Savitzky-Golay filter, and its first order derivative computed. Residues for which the first order derivative is positive are designated as natively unstructured, whereas those with negative values are structured.
Name Annotation type Description
Disordered Region Sequence Feature
Sequence features marking range(s) of residues with positive dydx values (correspond to the #Disorder column from JABAWS results)
Globular Domain Sequence Feature Putative globular domains
Dydx Annotation row First order derivative of smoothed score. Values above 0 indicates residue is disordered.
Smoothed Score
Raw Score
Annotation Row The smoothed and raw scores used to create the differential signal that indicates the presence of unstructured regions.
These are hidden by default, but can be shown by right-clicking on the alignment annotation panel and selecting Show hidden annotation

Documentation and thresholds for the JABAWS Disorder predictors adapted from a personal communication by Nancy Giang, 2012.