Curve fitting

Curve fitting

After generating a sensorgram, the next step is fitting the curves. The first and only model to use is the 1:1 Langmuir interaction describing the single exponential of the data. With the first attempt, leave the bulk contribution and drift on zeor and do not fit. Start by providing the initial fitting values and press “fit” to see the result. If the sensorgram consists of nice curves the fit will be ready in seconds; the fitted line follows the curves and will procide well defined values with small errors.

Fitting curves
1:1 interaction fitting
Fitted values
Fitted values

More realistic, optimization of the initial values and adjusting the fitting ranges may be necessary. This is normal and it is a good idea to fit the sensorgram with several initial values that differ a magnitude in scale. If, after fitting with different initial values, the results converge to the same values, it can be assumed that the values are well defined and robust. After the initial fittings, bulk contribution and/or drift can be fitted when necessary. Both bulk concribution and drift should be small compared to the curve response and proportional to the analyte concentration.

The next step is to critically compare the fitting with the measured curves. Does the fitting follow the curves? Are the calculated buffer jump values in line with the curves? Are the values of the parameters (Rmax, ka, kd) possible? Some instruments provide a quality control on the fitted parameters, however, always inspect the parameter values and the curves yourself.

Large RI jumps
1:1 fitting with large RI jumps

In the figure ‘1:1 fit with large RI jumps‘, the fit is not following the curves during the dissociation phase. By adding a larger buffer jump the program attempts to make the fit better. One other thing immediately apparent is that the overall response is too high. A way to solve this is to immobilize less ligand and repeat the experiment.

Other small problems, like small buffer jumps or low baseline drift can be solved with subtracting reference channels and blank (buffer only) injections (double referencing).

To validate further the interaction, several more steps are necessary. For instance, reverse the ligand and analyte and perform the experiments again. If you used the kinetic protocol, try an equilibrium experiment. Repeat the experiment with different batches of the same reagents to identify batch-to-batch variation. In all cases, a same value for kinetic constants is expected.

Read more

Read more about data fitting at the Data fitting introduction page.


# Lei 2016-03-16 07:29
How could I do the fit by adding a larger buffer jump? Thank you.
Reply | Reply with quote | Quote
# Arnoud1 2016-03-16 19:28
You can set the RI in the fitting model to a constant value.
Reply | Reply with quote | Quote
# Omar 2016-01-18 12:33
Please, what can I if I have mass transfer curve [I have like calibration curve]
Reply | Reply with quote | Quote
# Arnoud1 2016-01-18 19:14
You can incorporate a mass transfer term in the fitting.
Look at

Better is to avoid mass-transfer by immobilizing less ligand
Reply | Reply with quote | Quote
# ahsaab1976 2015-11-18 02:02
Hi Arnoud,

When I run the fitting in BiaEvaluation I always get sharp lines, I mean the lines don't go smoothly with the curve, always go with right angles around the curve without any curvature, do you why and how can I resolve this
Reply | Reply with quote | Quote
# Arnoud1 2015-11-18 11:06
You can try to change the initial values of the fitting. And start by fixing the RI in zero.

You can also post a question at the forum and include a sensorgram for clarification
Reply | Reply with quote | Quote
# Vivian Huerta 2013-10-15 21:32
Hi Arnoud:
What if the Rmax value is the one that does not seem possible? For example if it looks way too high? I am facing this problem in my experiments, I have tried to use a constant Rmax in the fitting which I have estimated experimentally and the KD varies in from 10e-09 to 10e-10
thanks in advance for your comments
Reply | Reply with quote | Quote
# Arnoud 2013-10-15 21:33
Did you try to fit the curves individually to get an idea what can cause this strange behaviour?
Reply | Reply with quote | Quote
# Jon Ashley 2012-11-19 20:34
Does this apply to concentration assays? I think a distinction has to be made between the quailty of kinetic based assays and concentration based assays somewhere.
Reply | Reply with quote | Quote
# Arnoud 2012-11-19 20:35
The focus of this page is on the kinetics.
Reply | Reply with quote | Quote