An ELISA was performed to detect an analyte. Now the question is IF and HOW the results can be used since different analyte concentrations were obtained when using different dilutions. No linearity is given for the samples. Samples above and below the LOD were already eliminated. For more information, see the attached file.
I am helpful if you can give me any advice how to find out which concentrations my sample have, when no linearity is given.
First of all, the standard curve should plotted with concentration on the x-axis (independent variable) and OD on the y-axis (dependent variable).
Data having very low OD measurements, close to the zero on the standard curve, should not be used.
Data having very high OD measurements, greater than about 2, should not be used because the standard curve bends over sharply, so that the accuracy is diminished.
After excluding those data, you could average the results for the remaining dilutions. If necessary, you can do a statistical test for outliers to see if they should be rejected, such as the Q-test.
https://en.wikipedia.org/wiki/Dixon%27s_Q_test
Thank you for your answer, Adam. This is clear for me and I already excluded values that are not in the acceptance range. What my question though is, is what to do when I get the following results for example:
O.D.: 1,572 concentration (interpolated): 336 dilution factor: 1
corrected concentration for dilution factor: 336
O.D.: 0,814 concentration (interpolated): 87 dilution factor: 2
corrected concentration for dilution factor: 174
O.D.: 0,541 concentration (interpolated): 47 dilution factor: 4
corrected concentration for dilution factor: 189
O.D.: 0,13 concentration (interpolated): 24 dilution factor: 16
corrected concentration for dilution factor: 387
As you can see, the results are completely different depending on which dilution I use. I would be interested if there is a method how to determine, which values are valid or if I need to choose another method in that case for analysing the concentrations.
I think you need to be clear about what is being excluded. I read it that SAMPLES giving values outside the acceptance range were excluded) i.e., very low and very high values). However, I think the issue here is your standard curve. I don't fully understand the way you have expressed the results, but it seems to me that you are using a standard curve which itself includes a value outside the 'safe' range (most people would worry about values above 2, unless you have strong data to suggest extended linearity of your assay).
You actually have a near linear plot if you omit the point at OD ~2.5. This high point has caused you to have a curved fit, when all other data points between OD 0 and 1.6 suggest that a linear fit is best. I'm not sure how this will affect your results, but I'd reverse the axes as advised by Adam B Shapiro and fit to y =mx + c. Given that your background is 0.0425 and the (new) high point on a linear plot will be 1.6, you should only interpolate sample values that fall between around 0.2 and 1.6.
My feeling is that your measurements at different dilutions, excluding those outside the dynamic range of the standard curve, are actually not that far apart. It may only take some minor improvements in technique to reduce the spread of values.
Nick Gee and Adam B Shapiro thanks a lot for that interesting input!
I could not find anywhere information about values above 2 being outside the "safe" range. Can you send me a reference for that info?
I though that I could use all values of my standard curve when using a 4-PL curve-fit like suggested in the manual. So if I understood you correctly, I need to determine the linear range within my standard curve, which I created with a four parameter logistic curve-fit? If so, how do I determine the linear range of my standard curve, when my standard curve was made with 4-PL?
I don't know of a reference as such, but it is just an observation that most standard curves tail off when you move above OD 2. This can be a feature/limitation of the instrument or the assay, or both. If you imagine your plot the conventional way round (y =absorbance), your own data clearly shows this effect. If you pushed this even further, small changes in absorbance would lead to huge changes in interpolated concentration, as the curve progressively flattens (or in your case accelerates upwards).
You seem to have a roughly linear response up to OD 1.6 though the points are not ideally spaced, and I would insert one or two points between 0.59 and 1.60. The linear range is simply the region where if you double (or in your case triple) the concentration, the absorbance will do likewise. In this case of course, the line of closest fit to the points is straight.
I would construct the standard curve to give more evenly spaced points on the y axis. If you are lucky, you will get a plot which is approximately linear and thus very simple mathematically. However, should the data show pronounced curvature, you will need to use another plot that fits better to the actual points. On the basis of your present data, I don't think that 4PL is the best fit as your standard curve does not seem to be sigmoidal. I would guess the best fit is linear (up to 1.6 OD), and possibly hyperbolic if you prefer to use all the data points. When curves have definite sigmoidal character (not evident in the data you shared) 4PL or 5PL may give the best fit.
- Badges
- Science topic
- Similar topics
- Biological Science
- Immunology
- Antibodies