Ubiquitin ΔΔG predictions


Select a prediction run

Note: Some functionality is currently disabled. Please contact Shane.OConnor@ucsf.edu to enable it again.

Methods

The original structures above contained the sequence for human ubiquitin, excepting the SH3 complex which contains the yeast sequence. There are three mutations between human ubiquitin and yeast ubiquitin: P19S, E24D, A28S. To treat the original structures as yeast ubiquitin, these three mutations were made to the original PDB structures using a fixed backbone mutagenesis protocol (fixbb) before ΔΔG prediction.

Structural predictions were made using the ddg_monomer Rosetta protocol using the command lines from row 16 of the table in the associated publication. These structures were then rescored by Kyle Barlow to determine binding affinity predictions.

Before computational methods were applied, solvent, ligands, unnatural amino acids and any superfluous chains in the asymmetric unit were removed from the structures. For NMR structures, only the first model was kept. Ubiquitin was set to chain A in the models.

Results files

The results of each prediction run can be downloaded by clicking on the , {}, and buttons above to download the results in TSV, JSON, or pandas format respectively.

All sets of results contain the same relevant information but each is better suited for a particular use: TSV files may be loaded directly into a spreadsheet program; JSON files can be used directly in Python or JavaScript; and pandas dataframes can be used directly in Python or used to pass from Python to R.

All sets of results may be loaded into R for analysis. Pandas dataframes may be loading using rpy2 (see here).

JSON example

To print the output in Python:

import json datapoints = json.loads(open('ubiquitin.json').read())['data'] for d in datapoints: print('%(PDB)s, %(Mutation)s, %(ddg_affinity)f' % d)

pandas example

To print the output in Python:

import pandas dataframe = pandas.read_hdf('2y2W9N.hdf5', 'dataframe') print(dataframe)

Note: the column 'Predicted' in the dataframes contains the binding affinity values.

Website issues / contact

Please contact support@kortemmelab.ucsf.edu if there are any problems with the website. Also, if you have any ideas on how to improve the presentation of the data, please let us know.