A certain curvature is normal. You do not show any figure of a typical standard curve you obtain; so it is difficult to judge it.

There could be an effect of the buffer you use to dissolve the BSA or use for the final dilution to 20uL. Bradford reagent is acidic (contains phosphoric acid). If your buffer is too strong and/or too alkaline, you should expect some interference.

Try increasing the volume of reagent, e.g., double its volume of Bradford reagent in your assay or go even higher, but keep the volume and concentration of the BSA the same.

A curvilinear graph is more practical in estimating protein concentration by Bradford method although linear regression might seem more feasible. An R2 value upto 0.97 for linear regression is acceptable for Bradford reagent.

As Achim suggested, the buffer could play an important role in estimation of the protein. Be sure that your buffer doesn't contain any detergent.

Sometimes reaction surfaces also play a part in giving such kind of result variations. Therefore it is better to go with the method which gives a curvilinear graph.

Make sure that the temperature during reaction time for both methods doesn't vary much.

For the same protein in the same buffer, each time it is not required to take BSA standard measurements, but it might improve your results and can reflect the errors or contamination in your sample, if any.

If you change your buffer and you have a different protein to measure a concentration, you might want to repeat the BSA standard curve measurement.

In the microplate assay format with commercial Bradford reagent, I find that the BSA standard curve is linear up to about 6 µg. If I extend the curve to 11 µg, it is better fit using a 2nd-order polynomial. I would not go as high as 20 µg because the curve flattens out rather badly.

Hello Adam B Shapiro: I have never done the Bradford as microplate assay.

My advice to Vikki Wong to increase the reagent fraction in samples, would that not increase the linear range of the standard curve?

I have prepared a standard curve using commercially purchased BCA protein assay kit. I didn't follow bradford method for microplate assay though. Please find the protocol it might help you. Stndard BSA protein concentration was in range of 125 µg to 2000 µg per ml and the absorbance range was 0.072 to 1.134 . R square value was 0.997

Achim Recktenwald Increasing the Bradford dye to protein ratio probably would extend the linear dynamic range of the assay. An alternative to that is to prepare multiple dilutions of the samples so that some of them fall within the dynamic range of the standard curve. In the 96-well microplate format, it is easy to do a lot of assays, since the whole plate can be read at once, so it is feasible to prepare various sample dilutions. I use three rows for 12 standards in triplicate (either 0, 0.5, 1, ..., 5.5 µg [linear fit] or 0, 1, 2,..., 11 µg [parabolic fit]). The remaining 5 rows are available for up to 60 samples. I use 10 µl samples and 200 µl of Bradford rgt. An 8- or 12-channel pipettor makes it fast to add the dye.

The Bradford assay does not work well when your concentrations are this high. If you go much higher than 0.3mg/mL, it always gives a curvilinear response. You should make your standards lie between about 0.02mg/mL and 0.2 mg/mL, and serially dilute your samples so that you always have an actual concentration for the samples within this range.

If you switch to bicinchionic acid, you can go to higher concentrations, and it also circumvents the problem of a potential mismatch between using something like BSA as a standard and the proportion of cationic residues in your protein. However, bicinchionic acid reagent is unstable and expensive, so I do not favor it as much as I used to.

The BCA is, however, the better assay. It is more generally "protein-specific", less dependent on specific amino acids. With Bradford, the old Lowry and some other assays, you can easily have results that are skewed by >30% to 40%.

I recommend to check out the chemistry of the different protein assays, to know what their preference is and what effect this may have on your problem and the results.

I do agree that it is the better assay, Achim Recktenwald, particularly as regards whether your standard does or does not match the residue profile of your source protein. If you wonder what we are disputing, Vikki Wong, it is that the Bradford reagent complexes with the cationic residues of a protein. There are a certain proportion of such residues per gram of standard, and a proportion of such residues per gram of your protein. If the proportion is not the same - and in some proteins the difference can be as much as 30% to 40% - then the standard will not accurately reflect how much protein you have.

The bicinchionic acid reagent (BCA) reacts with peptide bonds, unfolding your protein in the process and complexing with every peptide in the backbone. For this reason, for any tolerably large protein, there can not be more than a 1% difference in the reported concentration of your standard and that of your protein. Also, BCA does not give much inner filter effect behavior, even at concentration of 1 mg/mL. Standards for BCA should range from 0.1mg/mL up to 1mg/mL.

My only objection to the BCA test is that you really do have to use the reagent quite quickly after you blend it or it is wasted, and that it is much more expensive than the Bradford reagent. If I am preparing 100 samples, I cannot seem to get more than 25 prepared before my BCA reagent has started to change color. I have to make multiple sets of standards, just to keep pace with the reagent decay.

Benjamin Lasseter : It's a while ago that I messed around in the lab, but I never had the problem that the BCA reagent changed color as fast as described by you. I would try a different supplier for the reagent; this does not sound right.

I must sadly declare that the direct utilization of Photometer and/or Nanodrop without purification is not quantitative at all. Thus, the high-recovery HPLC-photometric methods (please see files; "Lysozyme by RP-HPLC" and "HPLC-Surf-SEC protein determination method") and PDMD (Protein-Direct-Microsequencing-Deciphering) method (please see files; "HepG2 Fucoidan" and "JMBT Alopecia") uniquely give the true determinative results. The high-recovery HPLC-photometric methods uniquely give the linear regression or linear calibration lines inserting to zero point due to obeying the Lambert-Beer Law. Further, low-recovery MS and/or HPLC-MS are also not quantitative at all.