Ab Initio Structure Prediction
Protein structure prediction in cases where no suitable homologous protein structures can be identified and used as a starting point. This method can be applied to sequences of 150 or fewer residues. It is primarily used for protein design in combination with aggressive sequence design methods such as RelaxDesign. As an example, ab initio structure prediction is part of the design/test cycle used by the Baker lab in the most recent designs of small protein binders to protein targets. These methods were first described in Simons et al., Proteins, 1999. Rohl et al., Methods in Enzymology, 2004. Variants of these methods can incorporate sparse experimental data (for example mutational data) to build experimentally useful predicted structures. Workflows can be customized to use experimental data to fit a customer’s specific needs. “One contact for every twelve residues allows robust and accurate topology-level protein structure modeling.”
Kim et al. Proteins Vol. 82. 2014.
Antibody/Antigen Structure Prediction
This method utilizes antibody structure prediction, antigen structure prediction, docking and optimization of their interaction. The degree of challenge depends on homolog sequence identity and knowledge of binding area.
Antibody Structure Prediction
Antibody Structure Prediction is a version of RosettaAntibody described in Weitzner et al. (“Modeling and docking of antibody structures with Rosetta” Nature Protocols Vol. 12 p. 401 (2017)), hardened to perform better on arbitrary sequence input and stabilized to run to completion on many types of input. This tool predicts the structure of the Fv region of the antibody from sequence. The Cyrus version adds Machine Language and stably runs the vast majority of monoclonal antibody sequences to completion, giving useful outputs in all situations. It has been successfully validated in the AMA II benchmark set and runs with high throughput: 500 predictions/week at the lowest tier, and even more throughput at higher tiers. Perfect for bulk predictions of antibody structures from sequences coming from methods such as yeast surface display and Next Generation Sequencing.
Homology Modeling of a Protein/Protein Complex
This is an iterative version of the Hybridize protocol in Rosetta, which is used for single chain protein structure prediction. This approach is fundamental to the consistently leading performance of Rosetta in CASP and the weekly CAMEO competition for protein structure prediction. In this method, single chain structures are predicted using the standard hybridize protocol. Templates are then identified where multiple chains appear simultaneously. Such templates are used to orient the single chain hybridize outputs and ultimately to produce a model of the overall protein/protein complex structure. This method is particularly valuable when high resolution structures of a protein target are not available, as it leverages Rosetta to its maximum extent for such extremely challenging and valuable targets. This method is published in multiple papers. One example of its application is “Near-atomic model of microtubule-tau interactions”.
Kellogg et al Science vol. 360 p. 1242 (2018).
Homology Modeling of Globular Proteins
Homology Modeling w/ DNA/RNA
This is a version of the hybridize protocol in Rosetta, which is customized to include small molecules during the modeling process. The small molecules are automatically parameterized for use with the Rosetta force field, using a process that combines specialized Cyrus software, Rosetta utilities, and software developed by OpenEye. Geometric constraints from a user-specified template, describing distances between the ligands and their binding sites, are employed during later stages of the HM refinement, so that the final output models contain the desired ligands properly positioned. This is the recommended protocol in cases where the binding configurations of one or more ligands to a similar protein is known, and ensures that the output models reflect that binding. The improved reliability of the binding sites in the resulting models makes these models particularly well suited for use in docking and virtual screening experiments. This method is published in multiple papers. An example of its application is “Near-atomic model of microtubule-tau interactions”.
Kellogg et al Science vol. 360 p. 1242 (2018).
Homology Modeling w/ Small Molecules
This is a version of the hybridize protocol in Rosetta, which is customized to include small molecules during the modeling process. The small molecules are automatically parameterized for use with the Rosetta force field, using a process that combines specialized Cyrus software, Rosetta utilities, and software developed by OpenEye. Geometric constraints from a user-specified template, describing distances between the ligands and their binding sites, are employed during later stages of the HM refinement, so that the final output models contain the desired ligands properly positioned. This is the recommended protocol in cases where the binding configurations of one or more ligands to a similar protein is known, and ensures that the output models reflect that binding. The improved reliability of the binding sites in the resulting models makes these models particularly well suited for use in docking and virtual screening experiments. This method is published in multiple papers. An example of its application is “Near-atomic model of microtubule-tau interactions”.
Kellogg et al Science vol. 360 p. 1242 (2018).