Averaging (ave)

Take the average distance distribution over different atom types from experimental conformations as the reference state, which means the distance distributions for all types of atom pair are identical in the reference state [25].

Quasichemical approximation (kbp)

Use the overall distance distribution of atom pair from experimental structures and calculate the specific distance distribution of atom types i and j based on the mole fraction (on the whole dataset) of atom type i and j [26].

Finite idealgas (dfire)

Treat the reference state as finite idealgas that probability of atom pair in a particular distance bin increases in ra with a tobedetermined constant a (a < 2) [27].

Spherical noninteracting (dope)

Treat the reference state as a sphere in which all atoms of a protein evenly distributed without ineraction. The size of sphere is specifically decided by corresponding experimental structure [9].

Randomwalk chain (rw)

Treat the reference state as an ideal randomwalk chain of a rigid step length, which mimics well the generic entropic elasticity and inherent connectivity of polymer protein molecules and yet ignores the atomic interactions of amino acids [29].

Atomshuffled (srs)

Generate a shuffled structure dataset by preserving all atomic positions while shuffling atom identities within each of the experimental structures [28].
