Supported Algorithms

Disorder Atlas uses the intrinsic disorder prediction algorithms that have been listed below. Please note that this is merely a summary, and for a thorough description of each algorithm please visit the respective websites.

IUPred (Dosztanyi et al., 2005): predicts disorder by analyzed estimated pairwise amino acid energy content. IUPred is based on the observation that the residues in proteins exhibiting intrinsic disorder are less likely to form stabilizing interactions.

DisEMBL (Linding et al., 2003): a suite of three disorder prediction methods: (1) Coils, (2) Hotloops, and (3) Rem465.

  1. Coils: A method that utilizes secondary structure prediction to predict disorder. As described by Linding et al., 2003, alpha-helices, 310-helices, and beta-strands are ordered whereas all other states are considered coils. Coils provides an overestimate of disorder, as disorder is always found within coils but coils are not always disordered. For this reason, Disorder Atlas does not utilize coils predictions, but instead uses hotloops (described below).

  2. Hotloops: The subset of coils predictions with high mobility, predicted by analyzing coil regions containing amino acids with high alpha-carbon temperature factors (B-factors).

  3. Rem465: A neural network trained on missing electron density (REM465 entries) in the Protein Data Bank (PDB).

Other external tools:

CIDER (Holehouse et al., 2015): Classification of Intrinsically Disordered Ensemble Regions. CIDER calculates numerous parameters relevant to protein disorder, including hydropathy and charge distribution.