 |
|
IMP Tutorial
|
In many cases our data comes from a heterogeneous sample. Heterogeneity can be both compositional and conformational. Interpretation of data collected on a heterogeneous sample requires a multi-state model. A multi-state model is a model that specifies two or more co-existing structural states and values for other parameters, such as the weights of each state.
IMP includes a command line tool multi_foxs for enumeration and scoring of multi-state models against SAXS profiles. There is also a web interface available, which functions similarly.
In this example, we will demonstrate how to use it using PDB structures of replication protein A and a simulated SAXS profile.
This tutorial can be followed in several ways:
multifoxs.ipynb.For full help on the multi_foxs tool, run from a command line:
The structure of the RPA in complex with ssDNA is available in the Protein Data Bank (PDB) as code 1jmc (file 1jmc.pdb in the multi_foxs/rpa directory), and the unbound RPA structures as code 1fgu (file 1fguA.pdb, 1fguB.pdb), while the SAXS profile is given in the weighted.dat file. The SAXS profile was simulated from the 3 structures with the following weights:
3% Gaussian noise was added to the simulated profile.
The atomic structures can be fit against the SAXS profile by running MultiFoXS:
FoXS calculates the theoretical profiles of the input structures, enumerates multi-state models, and fits them to the input SAXS profile. The output is a list of best scoring multi-state models of size 1, 2, and 3 in files ensembles_size_1.txt, ensembles_size_2.txt, and ensembles_size_3.txt, respectively. It also generates multi_state_model_x_x_x.dat files, containing the computed theoretical profile for the multi-state model and its fit against the experimental profile.
In this example, the best fit is obtained using a 3-state model (𝝌=2.80) ensembles_size_3.txt, with weights of 0.53, 0.23, and 0.25, showing a maximal deviation of 2% compared to the weights that were used to simulate the profile:
