Supplement to ECCB 2010: A Convex Programming Model for Protein Structure Prediction

Chris Kauffman and George Karypis

This page contains data supplementary to the article submitted to ECCB2010.

Code

The code for Domo is in two pieces. A modified version of the SDP solver CSDP was used in Domo. It must be downloaded and compiled. The main code for Domo comprises a series of Matlab files which are in the below archive.

csdp.tgz
Modified version of CSDP. Unpack and compile. Make sure csdp/solver/csdp is in your shell PATH.
domo-0.1.0.tgz
The main Domo code. Unpack and follow the instructions in INSTALL. Some example data files are included.

Constraint Violations

Supplemental discussion on the constraint violations for each method is found in the following file

supplement.pdf
Discussion of constraints violations for each method

Data Files

We also provide the raw data that resulted from our large scale experiment. 124 proteins were used at 5 noise levels with 5 different perturbations for a total 3100 predictions per method. The data files for the predictions have the following format.

  1. Unique key identifying the protein, noise level, and which random perturbation was used.
  2. (rmsd) RMSD between the predicted structure and true structure
  3. (Minutes) Amount of time to make the prediction in minutes
  4. (corr) Correlation of perturbed and unperturbed distance matrix
  5. (noi) 1 - corr
  6. (iprod) Normalized inner product between perturbed and unperturbed distance matrix
  7. (nrm) Normalized Euclidean distance between perturbed and unperturbed distance matrices
  8. (nb) Number of balls (atoms) in the model for this protein
  9. (nr) Number of residues (amino acids) in the sequence for this protein
  10. (rinit) Which of the five random perturbations was used for this prediction
  11. (id) ASTRAL identifier for the protein
  12. (lvl) Noise level (gamma) for this prediction
  13. (host) Which compute node the prediction was made on
cmds.dat
Prediction statistics for the CMDS method.
nedm.dat
Prediction statistics for the NEDM method.
domo.dat
Prediction statistics for our Domo method.
objz1.dat
Prediction statistics for our Objz1 method.
objz2.dat
Prediction statistics for our Objz2 method.