Our delta delta G (DDG) tool has been significantly improved in order to make better predictions of the stability of single point mutations. This method has been significantly changed in order address certain weaknesses in the previous approach (details below).
Will this affect previous DDG results?
Previous DDG runs will still have the values reported with the original DDG tool. When looking at results from the original DDG, you will see this message:
For more accurate results, you may want to repeat the DDG with this improved version.
IMPROVEMENTS
The new tool, termed Cartesian DDG in the Rosetta community, has upgraded the following features…
- Improved Energy Function
- Improved Predictions for Mutations that Change Net Charge
- Improved Solvation
- Improved Modeling Around Prolines
- Increased Non-Bonded cutoff
- Improved Side Chain Optimization
- Improved Structure Preparation
- More Comprehensive Analysis
IMPROVED ENERGY FUNCTION
The new DDG method uses the latest Rosetta scoring function, which improves upon earlier functions through modulation of term weights and through the incorporation of additional terms to the function. The biggest changes are improved electrostatic and bonded parameters and enabling attractive forces for hydrogens. These changes were made possible after benchmarking with newer data sets. They also used improved calculation methods for measuring similarities between structure predictions and positive controls.
For more information on the new Energy Function:
Alford et al. The Rosetta All-Atom Energy Function for Macromolecular Modeling and Design. J Chem Theory Comput. 2017 Jun 13;13(6):3031-3048.
IMPROVED PREDICTIONS FOR MUTATIONS THAT CHANGE NET CHARGE:
The new method has significant improvements to calculations of electrostatic and partial charges. This improves structure prediction in many ways. It is particularly helpful in relieving the known issue of the previous version, which caused inaccuracies when making predictions about mutations that alter net charge.
IMPROVED SOLVATION:
An improved implicit aqueous solvent term represents one of the most significant improvements to the Rosetta scoring function. This score term uses the anisotropic polar solvation model. This adds a virtual water molecule at ideal locations for polar side chains. Then if another residue in the protein would interfere with that interaction, there is a minor score penalty. This is known as the cost of desolvation. This was under-represented in the previous score term. The Rosetta community calls this LK Ball and it only affects polar side chains.
IMPROVED MODELING AROUND PROLINES:
Mutation to and from proline will usually have a huge effect on backbone angles. The new version of DDG implements a new way of sampling local backbone changes in order to better identify favorable local backbone geometries.
INCREASED NON-BONDED CUTOFF:
The new approach increases the distance for attractive and repulsive forces affecting atoms from 6 to 9 Angstroms. This is found to provide meaningful improvement in calculations by capturing weaker forces that can have a strong cumulative effect.
IMPROVED SIDE CHAIN OPTIMIZATION:
Side chains are now repacked and minimized in cartesian space in order to better capture conformational changes that occur due to point mutations. This allows DDG to capture many minor conformational changes at the site of mutation that were previously missed.
IMPROVED STRUCTURE PREPARATION:
Input structures need to be fully optimized in order to reduce noise in the DDG results. The new version of DDG runs 20 repeats of Relax with the improved scoring method in order to find a local energy optimized input conformation. This preparation is done under the exact conditions used in the Cartesian DDG calculations, including the increased non-bonded cutoff distance, avoiding the issue in previous versions where additional optimization of the native structure was required around each mutation site. This allows for the use of a single structure as the reference for all mutations done for this target which provides more meaningful and consistent results when ranking different mutations by their change in energy.
MORE COMPREHENSIVE ANALYSIS
The new version of DDG performs up to 5 repetitions of the mutation in order to better identify a converged local energy minimum state of the mutated state. This reduces noise in the results due that can occur from modeling trajectories that don’t find the best energy minimum in proximity to the Relaxed structure. Most point mutations converge easily if the input structure is optimized and the mutation is modest, but more extreme side chain changes may need more sampling to get converged score.
How does the new method compare with previous DDG in terms of measurable results?
This method was extensively benchmarked by the creators of this tool with a set of proteins that has experimental stability information about the effects of mutation. Cyrus also benchmarked this method with a benchmark set that improved on it by selecting more even representation numbers of mutation types. Both showed that the new DDG version is more accurate, more consistent, and suffers from fewer outliers.
For more information on:
- the Energy Function click here.
- DDG click here.
- Cartesian DDG and energy function improvements are described in the following publication:
- Park H, Bradley P, Greisen P Jr, Liu Y, Mulligan VK, Kim DE, Baker D, DiMaio F. Simultaneous Optimization of Biomolecular Energy Functions on Features from Small Molecules and Macromolecules. J Chem Theory Comput. 2016 Dec 13;12(12):6201-6212.