In his capacity as Prime Product Manager, Tyler Day oversees and personally contributes to the development of Schrödinger’s program for homology modeling and protein structure prediction. In this article, Dr. Day describes what’s new in Prime 3.0, released as part of Schrödinger Suite 2011.
The recently released Schrödinger Suite 2011 marks the debut of Prime 3.0, which contains a number of significant advances in the field of protein modeling. Among the hallmark features of Prime 3.0 is VSGB2.0, a new implicit solvation model that represents a marked improvement over its predecessor. Additionally, Prime 3.0 contains numerous improvements that help to make homology models more accurate and easier to create.
The VSGB2.0 Solvation Model
At the core of Prime 3.0 is an entirely new solvation model, VSGB2.0. As with the solvation model previously employed by Prime,1 VSGB2.0 is based on an underlying surface-generalized Born (SGB) model enhanced with a variable dielectric correction.
The SGB model itself is an approximation to the exact Poisson-Boltzmann equation, representing the protein as a solvation surface immersed in continuum solvent. This choice of an implicit solvent model over an explicit one allows for the development and application of fast and highly efficient sampling algorithms such as that incorporated in Prime's loop prediction.
The variable dielectric correction approximates the energetic consequences of polarization effects. Such polarization effects, which are absent from fixed-charge force fields, are of key importance owing to the diverse electrostatic character exhibited in the protein environment. The variable dielectric correction accounts for these polarization effects by varying the internal dielectric constant used in the SGB model based on the chemical nature and identity of the amino acids under consideration at any given time.
VSGB2.0 then greatly expands on this variable dielectric model in several important ways. In addition to polarization, typical force fields and solvation models omit a number of other interaction types that drive protein energetics. To account for these various phenomena, the VSGB2.0 model incorporates additional physics-based terms designed to capture the effects of hydrogen bonding, pi-pi interactions, self-contact corrections, and hydrophobic packing.
Building homo- and hetero-multimers from the graphical interface
In addition to the VSGB2.0 solvation model, Prime 3.0 also introduces a number of significant enhancements to its homology modeling capabilities. Usability enhancements mean that users can now easily build homo- and hetero-multimers from Prime’s graphical interface. Now, when the user selects multiple chains as templates (in the Find Homologs step) and carries them forward (to the Build Structure step), the user can construct homo-multimers with a minimum of effort. Users of Prime 3.0 can, for example, construct homology models for tetrameric ion channels (e.g., hERG) as easily as they can construct a single chain model.
Similarly, hetero-multimers can be constructed by by using different runs within the Prime workflow, and combining them in the Build Structure step to construct the desired multimer.
In the Build Structure calculation the multimer’s individual chains are constructed and refined in the presence of each other, resulting in a self-consistent final model. This model is free of the clashes and other structural inconsistencies that arise when individual chains are modeled in isolation and subsequently recombined.
Pairwise constraints, structural feedback, and GPCR-specific settings
Prime 3.0 also introduces a number of other improvements designed to improve the quality of homology models that users can generate. Users may now define pairs of residues in the query sequence which are to be in close contact in the final built model. These pairwise constraints can be invaluable in ensuring that the built model satisfies known experimental contacts within the modeled protein. If the residues in question are cysteines, Prime will automatically construct a disulfide bridge in the final model.
Additionally, the homology models produced by Prime 3.0 are now colored according to the origin of each residue, indicating whether the coordinates for a particular residue were based on the corresponding coordinates from the template structure, or whether the residue was constructed ab initio during the model building process (see Figure 1). This provides immediate feedback on areas of high or low sequence identity, as well as regions where insertions were built or gaps were closed. It also provides a clear indication of what movement was required when satisfying any user-specified pairwise constraints.
Prime 3.0 also enhances support for building GPCR homology models. In recent months, several new GPCR crystal structures became available. These include two previously unsolved targets, CXCR4 and D3, as well as an activated form of the β2 adrenergic receptor. Prime includes annotated sequences for all GPCR structures published to date. Prime automatically uses these annotations for sequence alignment, ensuring that the well-known fingerprints for class A GPCRs are applied.
Enhanced robustness and improvements to associated programs
In addition to the scientific and functional advances, Prime 3.0 includes a complete overhaul of the underlying mechanisms for structure handling. This overhaul has removed Prime’s dependence on intermediate files and file formats, greatly increasing its ability to handle previously problematic structures.
Additionally, users of other programs that rely on Prime technology – including MM-GB/SA, IFD (flexible receptor docking), and covalent docking – can expect to share in the benefits of the new features mentioned above. Furthermore, many of these Prime-related programs have new features of their own, and their interfaces expose many new user-adjustable parameters.
Figure 1: A homology model of the activated form of the D3 dopamine receptor built on the activated form of the β2 adrenergic receptor (3P0G). Blue residues correspond to conserved positions in the alignment, with all heavy atom coordinates being derived from the template file. Cyan residues represent mutations, with backbone atomic coordinates derived from the template but sidechain atomic coordinates predicted ab initio. Red residues represent portions of the structure built entirely ab initio, in this case due to local insertions and deletions in the alignment.
Reference
(1) Zhu, K.; Shirts, M.R.; Friesner, R.A., J. Chem. Theory Comput., 2007, 3, 210.