I have a sigmoidal standard curve, and a similar curve for my sample, both generated from a 5-parameter logistic curve fitting. My question is, what is the standard/usual way to calculate the amount of my sample from the standard curve with this regression model?

Should I choose any ONE point (must it still be on the 'linear' range since I am not using the linear regression model; is the accuracy compromised if I chose those that are close to the upper and lower ends, but are still within the range of the standard curve?) on the sample curve, and find it's corresponding amount on the standard curve? And subsequently multiply the amount by the dilution factor corresponding to that point?

I am using the MasterPlex software to generate my curves. Does anyone know in what situation should I select the option 'Fixed lower asymptote zero'?

And if I do not have a standard curve, but would only like to compare qualitatively among samples, should I plot a linear-linear, linear-log or log-log curve? I know it doesn't really matter for qualitative comparison, but what is, again, the standard/usual way to analyse and present my data. I am studying Ig Isotype production from T-dependent antigen stimulation by the way. Some journal reference would be VERY MUCH appreciated.

Thanks in advance.

Submit your paper to J Biol Methods today!

# ELIZA

Started by yoshi, Nov 11 2010 07:00 AM

2 replies to this topic

### #1

Posted 11 November 2010 - 07:00 AM

### #2

Posted 16 November 2010 - 03:58 AM

here are some answers to your questions.

1. As you approach the low and high end of the dose response curve the %CV increases. If you were to plot %CV vs concentration you would have a plot that looks like a "U". So points in the linear portion of the curve are the most accurate. You can confirm this by running dilutions of your high samples so they fall in the linear portion of the curve. Also, running replicats of the low value samples or your 0 point you will be able to determine the %CV at the low end.

2. Standard curve v. sample curve. You are unclear on this...I am going to assume your sample curve is a curve of responses with values obtained from another methodology? In any case the sample values from a reference method should match your experimental assay method. Thus, the sample curve and standard curve should lie on top of one another. If this does not happen you have to adjust your conditions (pH, buffer, matrix) or your standards (calibrators? or matrix) to achieve comparable results.

1. As you approach the low and high end of the dose response curve the %CV increases. If you were to plot %CV vs concentration you would have a plot that looks like a "U". So points in the linear portion of the curve are the most accurate. You can confirm this by running dilutions of your high samples so they fall in the linear portion of the curve. Also, running replicats of the low value samples or your 0 point you will be able to determine the %CV at the low end.

2. Standard curve v. sample curve. You are unclear on this...I am going to assume your sample curve is a curve of responses with values obtained from another methodology? In any case the sample values from a reference method should match your experimental assay method. Thus, the sample curve and standard curve should lie on top of one another. If this does not happen you have to adjust your conditions (pH, buffer, matrix) or your standards (calibrators? or matrix) to achieve comparable results.

### #3

Posted 16 November 2010 - 04:02 AM

Additionally, with respect to dilutions for high samples you do multiply the result by the dilution factor. One way to tell the linearity of the assay is to run serial dilutions of a very high sample in your buffer matrix IN PARALLEL with the highest standard. Your sample should dilute linearly. If you see problems here you may have 'matrix effect'.