The DynaMine backbone and sidechain dynamics and conformational propensities are described in:

• From protein sequence to dynamics and disorder with DynaMine
  Nature Communications 4, 3741 (2013).
• The DynaMine webserver: Predicting protein dynamics from sequence
  Nucleic Acids Research 42, W264-W270 (2014).

The EFoldMine early folding predictions are described in:

• Exploring the Sequence-based Prediction of Folding Initiation Sites in Proteins
  Scientific Reports 7, 8826 (2017).

The DisoMine disorder predictions are described in:

• Prediction of disordered regions in proteins with recurrent Neural Networks and protein dynamics
  bioRxiv 2020.05.25.115253 (2020).

The Agmata beta-sheet aggregation predictions are described in:

• Accurate prediction of protein beta-aggregation with generalized statistical potentials
  Bioinformatics 36, 2076-2081 (2020).

With these predictions we try to capture the 'emergent' properties of the proteins, so the inherent biophysical propensities encoded in the sequence, rather than the behavior of a final folded state. This relevant as proteins are dynamic even when folded, and might not fold at all (as with intrinsically disordered proteins). Please see our website for more information on how to run these approaches on your own sequences.

These predictions are single-sequence based, and the median/quartile/outlier information in the plots is derived from a BLAST search of the protein against uniref90, using default Uniprot parameters, followed by the standard Uniprot alignment to obtain a multiple sequence alignment (MSA). We then run the predictions on each sequence separately, map them back to the MSA, and look at the 'biophysical variation' observed in evolution in relation to the residues of the original protein. These are not available for Agmata as the method is computationally too expensive.

The protein interaction predictions are from SeRenDIP, the epitope predictions from SeRenDIP-CE. Both methods use evolutionary information and were developed at the Vrije Universiteit Amsterdam in the group of Prof. Feenstra :

• Seeing the trees through the forest: sequence-based homo- and heteromeric protein-protein interaction sites prediction using random forest
  Bioinformatics 33, 1479–1487 (2017).
• SeRenDIP: SEquential REmasteriNg to DerIve profiles for fast and accurate predictions of PPI interface positions
  Bioinformatics 35, 4794–4796 (2019).
• SeRenDIP-CE: Sequence-based Interface Prediction for Conformational Epitopes
  bioRxiv (2020).
• SeRenDIP-CE: Sequence-based Interface Prediction for Conformational Epitopes
  Bioinformatics btab321, 1367-4803 (2021).

Planned:

Information from molecular dynamics trajectories, and NMR chemical shift derived data on the biophysical properties of experimentally studied molecule.