Fitting Models to Individual Experiments

Basic model pipeline

Under the hood, models of individual experiments do the following:

1. Guess model parameters.
2. Calculate the concentrations of all molecular species using guessed binding constants (as well as the titration shot sizes and initial concentrations in the cell and syringe).
3. Calculate the heat change per shot using the guess reaction enthalpies and species concentrations from (2).
4. Compare the calculated and measured heat changes at each shot.
5. Iterate through steps 2-4 using nonlinear regression to find parameter estimates.

Implemented Models

• Blank
• Single site
• Competitive ligand binding
• Binding polynomial
• Assembly autoinhibition

Example Fit

import pytc

# Create fitter
g = pytc.GlobalFit()

# Load experiments
a = pytc.ITCExperiment("demos/ca-edta/hepes-01.DH",pytc.indiv_models.SingleSite,shot_start=2)
g.add_experiment(a)

# Fit and show results
g.fit()
print(g.fit_as_csv)

Differences from other modeling approaches

pytc is implemented with the philosophy that one should model all processes in the ITC experiment and include them in the fit. These processes include not only the binding reaction of interest, but also systematic errors in sample preparation and heat arising from dilution. The latter effects are captured with nuisance parameters added to standard thermodynamic models of binding:

\[Q_{i,obs} = Q_{i}\alpha + D_{slope}[T]_{total} + D_{interecept} + \varepsilon\]

where \(Q_{i,obs}\) is the heat at injection \(i\) and \(Q_{i}\) is the heat at injection \(i\) arising from the reaction of interest. \(Q_{i,obs}\) differs from \(Q_{i}\) because of systematic concentration errors (\(\alpha\)) and the heat arising from dilution (\(D_{slope}\) and \(D_{intercept}\)). \(\varepsilon\) is the fit residual.

\(\alpha\) scales all heats, accounting for systematic errors in the relative protein or ligand concentration. This could also be viewed as an apparent stoichiometry but, following Freire et al., we interpret this as “the effective amount of active protein relative to the nominal value entered as protein concentration.” This term is referred to as the fx_competent (fraction competent) in pytc output. Practically, this value should be near 1.0. Large deviations from 1.0 may indicate that the model used does not describe the stoichiometry of the protein or that there is a problem with the experiment.

The other nuisance parameters, \(D_{slope}\) and \(D_{intercept}\), model the heat of dilution as a linear function of titrant concentration. This is equivalent to fitting a straight line through a blank titration and then subtracting that line from an experimental titration. Rather than subtracting a blank, pytc allows a user to incorporate a blank titration into the global fit: the user fits the same \(D_{slope}\) and \(D_{intercept}\) across both an experimental and blank titration, thus explicitly modeling the heat of dilution.

If a one sets the values of \(\alpha\) to 1.0, \(D_{slope}\) to 0.0, and \(D_{intercept}\) to 0.0, and then does not allow them to vary during the fit, one recovers only the thermodynamic model—as is done in software such as Origin.