Features Analysis Benchmark Protocol Capture

Overview

High-resolution crystal structures are assumed to be thermodynamically stable, which means that each structure should occupy a minimum of a high quality energy function. In particular, in aggregate, Rosetta should reproduce local structure features observed in the high quality crystal structures. Therefore if we relax native crystal structures and observe systematic displacement of local geometric features (specific types of bonds are lost, polar contacts become too close, etc.) this indicates that when new structures are predicted, they will likely not be stable in the lab.

In our 2015 H-bond paper, we observed that the interaction between related terms in the energy function create areas of double counting for certain types of interactions or contexts that may be maybe difficult to observe in full structure benchmarks. By plotting and looking closely at distributions of local features, it is possible to gain a deeper understanding of the full energy that arises from the complex interaction between the range of active terms and the implicit kinematic constraints.

Literature references

Combined Covalent-Electrostatic Model of Hydrogen Bonding Improves Structure Prediction with Rosetta Matthew J. O’Meara, Andrew Leaver-Fay, Michael D. Tyka, Amelie Stein, Kevin Houlihan, Frank DiMaio, Philip Bradley, Tanja Kortemme, David Baker, Jack Snoeyink, and Brian Kuhlman J. Chem. Theory Comput., 2015, 11 (2), pp 609–622 DOI: 10.1021/ct500864r

Scientific benchmarks for guiding macromolecular energy function improvement Andrew Leaver-Fay, Matthew J O’Meara, Mike Tyka, Ron Jacak, Yifan Song, Elizabeth H Kellogg, James Thompson, Ian W Davis, Roland A Pache, Sergey Lyskov, Jeffrey J Gray, Tanja Kortemme, Jane S Richardson, James J Havranek, Jack Snoeyink, David Baker, Brian Kuhlman Methods in enzymology 523 2013

Usage:

To install this package, in R:

if (packageVersion("devtools") < 1.6) {
  install.packages("devtools")
}
devtools::install_github("momeara/RosettaFeatures")

Generate features databases following the features_benchmark protocol capture

https://github.com/RosettaCommons/demos/tree/master/protocol_capture/features_benchmark/README.md

Then to report features, in R:

library(RosettaFeatures)
compare_sample_sources(
  config_filename="analysis_configuration.json")

Where the analysis_configuration.json looks like:

{
  "output_dir" : "native_vs_relax_native",

  "sample_sources" : [{
    "database_path" : "native_features/features.db3",
    "id" : "Native",
    "reference" : true
  }, {
    "database_path" : "relax_native_features/features.db3",
    "id" : "talaris2014",
    "reference" : false
  }],

  "analysis_scripts" : [
    "scripts/analysis/plots/EXAMPLE_PLOT.R"
  ],

  "output_formats" : [
    "output_print_pdf"
  ]
}