Programmatic flow of qFit protein algorithm.
A. qFit residue algorithm, demonstrated by Tyr118 in the E46Q mutant structure of the photoactive yellow protein from Halorhodospira halophila (PDB: 1OTA)29. The 2mFoCDFc composite omit density map contoured at 1 σ is shown as a blue mesh.
A.1. Backbone sampling: For each residue, qFit performs a collective translation of backbone atom (N, C, Cα, O) coordinates.
A.2. Aromatic angle sampling: For aromatic residues (His, Tyr, Phe, Trp), qFit takes the conformations from the backbone step and samples the Cα-Cβ-Cγ angle.
A.3. Dihedral angle sampling: Since Tyr has two χ angles, qFit starts by taking the output conformers from the aromatic angle sampling step and exhaustively samples the χ1 angle, scoring the best conformations based on QP/B-factor/MIQP scoring. qFit then uses these best conformations as input to sample the remaining χ angles in the Tyr residue. Since the only angle left to be sampled is the χ2 angle, qFit rotates about the terminal ring of the Tyr and then scores the conformations that best fit the density.
A.4. Final qFit residue scoring: Once we reach the terminal ring (all sampling steps have occurred), we perform QP and B-factor sampling, followed by MIQP with BIC selection. MIQP with BIC selection removes a redundant overlapping conformation, resulting in two distinct conformations of this Tyr residue. This model is then output as the residue multiconformer.
B. qFit segment algorithm, demonstrated by Tyr118 in PDB: 1OTA. After identifying all optimal conformations for each individual residue, qFit works to connect the protein back together.
B.1 qFit segment: Moving linearly along the protein sequence, qFit identifies ’segments’ of residues with multiple backbone conformations. Here, Ser117 (i) and Tyr118 (i+1) have multiple backbone conformations. qFit segment enumerates each possible combination of alternate conformations between these two residues, creating four possible combinations. The optimal combination of conformations is then determined by the QP/MIQP scoring, leading to one combination being culled.
B.2 qFit relabel: qFit uses Monte Carlo optimization with a steric model to assign altloc labels to spatially coupled alternative conformers. In this example, Ser117 and the neighboring Gln32 initially have clashing altloc B conformers. However, relabeling swaps the A and B labels of Gln32 to relieve this clash.
B.3 qFit refinement: We then refine the occupancies, coordinates, and B-factors of the raw qFit output file to produce a final qFit model.