ANCHOR predicts binding regions located in disordered proteins from the amino acid sequence. These regions function via undergoing a disorder-to-order transition upon binding to a globular protein partner. ANCHOR seeks to identify segments in a generally disordered region that cannot form enough favorable intrachain interactions, however they have the capability to energetically gain by interacting with a globular partner protein. Our prediction is based on three properties.

• The first criterion ensures that a given residue belongs to a long disordered region, and filters out globular domains.
• The second criterion corresponds to the isolated state and it ensures that a residue is not able to form enough favorable contacts with its own local sequential neighbors to fold, otherwise it would be prone to adopt a well defined structure on its own.
• The third criterion tests the feasibility that a given residue can form enough favorable interactions with globular proteins upon binding. This basically ensures that there is an energy gain by interacting with globular regions.

The prediction of these three properties relies on the energy estimation framework implemented in IUPred, a general disorder prediction method. The core element of IUPred is the energy predictor matrix P. The parameters in Pij were trained on globular proteins with known structures only, without relying on any kind of disordered dataset. These parameters were determined to minimize the difference between the estimated energies and the energies calculated from the known structures on the dataset of globular proteins.