Hey Asmaa,

this sounds like an assay problem. If your standard curve is not linear you can not calculate the concentrations. However, if the OD of your samples drop below your lowest standard (or your blank OD value) then the concentration of the substance you want to detect is to low for this assay.

Personally, I would not believe only OD values for interpretation of substance levels (for example cytokines). You perform an ELISA to dertermine concentrations, so I would highly recommend to do the assay again and get your standard curve right. I would calculate the CV based on the concentration replicates. I know its frustrating not to use the data already gathered, but I myself would not trust the results, if the assay looks strange from the beginning.

All the best and a nice day,

Marc

The Coefficient of Variation (%CV) should be calculated using the Standard Deviation of the Optical Density (OD) across each set of replicates. This measure is used to inform the experimenter conducting the assay as to the performance of the assay and whether any significant inconsistencies exist across replicates that would require the assay to be re-administered (generally anything below 15% is acceptable, unless you are regulated by FDA or similar agency). The curve only has to be linear within the range of your unknown (ie. you have standards plotted above and below your unknown such that the linear portion of the curve traverses through the unknown point), in general you can truncate the high-end or low-end values or use a polynomial curve fitting (if your CV/RSD comes into better alignment). You usually don't report CVs for any figures that arose as the result of calculations.

This might be useful https://www.rndsystems.com/resources/how-to-analyze-elisa-data

Dear Asmaa, CVs are calculated by Std Dev of ODs, as stated above, but do not forget to subtract the blank ODs from samples and controls, especially you are running ELISA on single filter plate reader. Since you are finding intra-assay CVs, it must be less than 10% overall. The CV% of duplicates or triplicates must be less than 20%, and it is recommended that as a better practice CV% of replicates should ideally be controlled less than 8%.

Outliers of standards can and cannot be taken out. Like in certain cases we cannot determine that an outlier is simply the wrong input data or could be something interesting or even new for that assay being tested. It can be taken out if it is due to some incorrect data entry or if does not affect or change the expected results of your assay. Generally, it is not acceptable to omit any deviated observation from data by just calling it as an outlier. You can run the analysis with and without that outlier and see the change in your result.

Regards...