Skip to main content

Parameter Files

Overview​

Parameter files contain initial estimates for the parameters used in the fitting process. These files are specified in ChemEx using the -p or --parameters option:

chemex fit [...] -p <parameter_file> [...]

Parameter files are organized into sections:

  • The [GLOBAL] section applies parameters universally to all residues.
  • Residue-specific parameters are defined in sections named after each parameter, such as [CS_A]. Multiple parameter files can be provided if needed.
warning

To ensure accurate results and avoid local minima, set appropriate initial values for each parameter, as the χ2 minimization process involves multidimensional searching.

info

If no initial value is provided in the parameter files, a default value will be assigned.

Example Parameter File​

Below is an example of a parameter file:

parameters.toml
[GLOBAL]
PB     = 0.015
KEX_AB = 70.0
TAUC_A = 10.0

[CS_A]
13N = 108.207
26N = 115.711
28N = 113.882
29N = 115.318
33N = 115.636
37N = 116.159
41N = 114.635
42N = 113.525
43N = 108.876
50N = 107.855
52N = 111.358
55N = 128.301
59N = 116.388
66N = 119.429
67N = 114.454
68N = 120.595

[DW_AB]
13N = 4.0
26N = 5.5
28N = 6.5
29N = 6.0
33N = 4.5
37N = 5.0
41N = 6.0
42N = 6.0
43N = 12.5
50N = 8.0
52N = 8.5
55N = -6.5
59N = 6.5
66N = 4.0
67N = 8.0
68N = 4.5
tip

Setting model-free parameters (e.g., TAUC_A) can provide a good initial estimate for relaxation parameters (e.g., R1_A, R2_A). For biomolecules in H2O at T = 300 K, the overall tumbling time is roughly 1 ns per 2.6 kDa of molecular weight. Tumbling time is proportional to η/T, where η is solution viscosity and T is temperature in Kelvin.

Setting Parameter Bounds​

To set upper and lower bounds for any fitting parameter, replace the initial value with a list of three elements:

parameters.toml
PARAMETER_WITH_NO_BOUNDS = <initial_value>
PARAMETER_WITH_BOUNDS = [<initial_value>, <lower_bound>, <upper_bound>]

Setting bounds helps prevent parameters from reaching unrealistic values during χ2 minimization. However, avoid overly strict bounds as they can hinder convergence. Some minimization algorithms, such as differential_evolution, require finite bounds for all fitted parameters.