GE Healthcare Life Sciences
Biacore Life Sciences
You are here: Biacore Life Sciences > Help > Kinetic evaluation - Inspect the fit
printer friendly

Kinetic evaluation - Inspect the fit

There are several numerical and graphical aids to judge how well the chosen model fits the experimental data. However, with a little practice, visual inspection of the fitted curves overlaid on the sensorgrams is often the best way to judge the results. Residual plots showing the difference between experimental and calculated data can be helpful.
In general, residual values will be larger for global fitting to all curves simultaneously than for local fitting to each curve individually. This is because global fitting constrains certain parameters to have the same value for all curves.
In an excellent fit, the fitted curves correspond closely to the sensorgrams over the whole range. The model is a fully acceptable description of the experimental data. Excellent fits are rare unless a major effort has been made in sample preparation and experimental execution.
Example of a good fit
In a good fit, the fitted curves correspond reasonably well to the sensorgrams. There may be minor discrepancies in the response levels reached or in the shapes of the curves, but these are small in relation to the overall response levels. Whether the fit is acceptable or not must be judged in relation to the purpose of the experiment. Experimental variations and unavoidable deviations from ideal behaviour in biological systems often mean that a fairly good fit is the best that can be obtained.
Visual inspection of the residual plot or of the fitted curves overlaid on the experimental data gives an indication of the closeness of fit. Ideally, the residuals will scatter randomly around zero over a range that corresponds to the short-term noise in the detection system. Deviations from ideal fitting appear as systematic variations in the residuals, imparting a non-linear shape to the residual plot. Judge the residual range and shape in proportion to the response ranges in the experimental sensorgrams and in relation the goal of the investigation.
Example of a poor fit
In a poor fit, the fitted curves deviate markedly from the sensorgrams, in the response levels reached and/or the shape of the curves. In particular, fitted curves that show systematic deviations in shape from the experimental sensorgrams usually indicate that the chosen model is not suitable. Reliable kinetic constants cannot be obtained if the fit is poor.

Statistical indicators

Some statistical information provided with the results can help you to judge the fit (all parameters may not be available in all systems):

Chi2 (χ2) is a measure of the average deviation of the experimental data from the fitted curve. Lower Chi2 values indicate a better fit. The acceptable level for Chi2 must be assessed in relation to the measured binding levels.
The Chi2 value in an ideal situation will be approximate to the square of the short-term noise level. It is however difficult to recommend absolute values for acceptance limits for Chi2: the values need to be considered from case to case in combination with assessment of the shape of the residuals (Chi2 is related to the overall range of the residuals but is not affected by the shape of the residual curve).
Standard error is a measure of the confidence in the reported value for the parameter. A small standard error indicates that changes in the parameter's value have a significant effect on the fitting: in other words, confidence in the value is high.
T-values are obtained by dividing the parameter's value by the standard error, and thus provide a kind of normalised inverse standard error value. High T-values (typically above 10) indicate confidence in the parameter's value.
U-value is uniqueness value for the kinetic rate constants. Lower values indicate greater confidence in the results.