I am very sorry for hearing your experiences in ELISA study.  

ELISA is not the quantitative method at all; i.e., (1) IgG (dimer) is not so specific as compared to IgM (pentamer and/or hexamer) (more than 10-fold increase in specific affinity is expected to be obtainable in IgM as compared to IgG; please see file "The Fascio effect") and (2) photometric assay without prior-purification beforehand gives notorious bent-calibration curve to give c.a. 2.5-fold lower determination values of fucoidan transported than HPLC method due to violence to the "Lambert-Beer law" (please see Fig. 10 in the file; SEC fucoidan determination).  Therefore, please use of the quantitative PDMD (Protein-Direct-Microsequencing-Deciphering) method instead of unreliable ELISA method to obtain the true quantitative values (please see file; HepG2 fucoidan and JMBT Alopecia).

By the way, human serum biotinidase is present at median 0.25 mg/mL (range; 0.19-1.02 mg/mL; n=13) (please see file; JMBT Alopecia).   Human blood contains serotransferrin at median 0.35 mg/mL (range; 0.0-0.72 mg/mL; n=13).   Human blood contains ceruloplasmin at median 0.36 mg/mL (range; 0.11-0.88 mg/mL; n=13).    Human blood contains albumin at median 33.9 mg/mL (range; 15.2-48.2 mg/mL; n=13).   Adult human blood (n=4; median32.5y, range 22-52y) contains connectin/titin at median 2.63  mg/mL (range; 1.55-4.08 mg/mL; n=4).   Children's blood (n=9; median 12mo, range 4-48mo)  has no  connectin/titin.   Human serum biotinidase is major protein of human serum as firstly pointed out by Dr. Krishnamurti Dakshinamurti (Emeritus Professor of Biochemistry & Molecular Biology, University of Manitoba, Department of Biochemistry and Medical Genetics, Winnipeg, Manitoba, Canada). 

Serum total protein (TP) as determined by HPLC-Surf-SEC method is median 67.4 mg/mL (range; 57.2-92.4 mg/mL; n=13).   Then, 270-fold purification is sufficient for serum biotinidase isolation.   However, low-recovery purification due to adsorption of hydrophobic biotinidase to the chromatographic gel-materials has given the reportedly 30,000-fold purification (which has suggested strange low-concentrations of the human serum biotinidase in human blood). 

Thus, the high-recovery purification is an important issue (please see file; HPLC--SEC BIN NP-40 and HPLC-Surf-SEC protein determination method).

Therefore, I would like to recommend to develop the reliable RP-HPLC determining method (please see file; Lysozyme by RP-HPLC).

First of all, your calibration curves are really extraordinary. This shape is an indication for a complex binding regime. This can be caused by the polyclonality of the capture antibody or something else. Usually, 4-parametric curves are not "interpreted" mechanistically - they are simple empirical functions, which work quite well if you have a simple system, e.g. based on monoclonal antibodies.

If you are interested in a quantitative determination of your analyte, you can use any function you like. Even a simple interpolation or smoothing of your data points is adequate for this purpose.

For a mechanistic interpretation, you need different functions, which for example are based on the law of mass action. Unfortunately, in most cases, a simple calibration curve is not sufficient to make a valid deduction of a mechanism, if this should be your primary interest.

Thank you both for your input.

Hi Kuo,

Thanks for all the publications. You've at least filled my day with reading :)

As for some of the comments, we have quantitative methods available to us, but want a faster one. Our determinations of plasma will be run against a purified standard. The plasma will be diluted >1000x, so I don't expect much interference from even the abundant plasma proteins. If I space out my ELISA addition steps, any biotinidase will be washed away before biotinylated protein is added to the plate, but thanks for bringing this to my attention. I was unaware of its existence. I'll let you know if I have more questions after reading some 

Hi Michael,

I see similar behavior with monoclonal captures and this polyclonal-equivalent detection protein, but I don't think that changes your statement.

I think you're right that I'm asking too much of this data and need to work on another experiment to tease out what is behind this curve. Regarding a mass transfer model, if you know of any appropriate to ELISA, let me know. The Langmuir isotherm cannot describe even my well-behaved ELISA's, and summed Langmuir's are similarly insufficient for the data I showed above. The linked paper describes 1:1 binding with a distribution of affinities, and that's not a bad thought. I'll give it some time and see if ELISA's appear to justify its use.

http://www.cell.com/biophysj/pdf/S0006-3495(03)75132-7.pdf

Amended on 12 September 2017

Furthermore, I would like to express on the ELISA issue.

I must sadly repeat about the differences in determination value between HPLC-photometric method and ELISA method (please see files to compare; SEC fucoidan determination and ELISA fucoidan).

The determined value of fucoidan (Okinawa-fucan) is c.a. 2,000-fold higher in the HPLC-photometric analysis as compared to notorious ELISA method (our experience together with Dr. Takeaki Nagamine, M.D., Graduate School of Health Sciences, Gunma University, 3-39-15 Showa-machi, Maebashi, Gunma 371-8514, Japan).   Dr. Nagamine has sadly thrown away the fucoidan-assay kit of ELISA when he is retiring from the University (my unpublished observation or experience).

(1) Direct photometric assay without purification gives only 50% value as compared to HPLC-photometric assay due to violence to Lambert-Beer's law (please see Fig. 10 of file "SEC fucoidan determination" again). 

(2) Further, binding protein is not so specific; i.e., avidin binds biotin (vitamin H), lipoic acid (thioctic acid), and amino acids, and binding is stronger to lipoic acid than biotin (please see file; Thioctic acid determination) .   Then, biotin assay by using avidin is very difficult since serum has 10-fold higher concentration of lipoic acid than biotin; i.e., lipoic acid may interfere the biotin binding reaction of avidin.   Similarly, binding protein of IgG seems to have a broad binding specificity. 

(3) Furthermore, specificity of binding protein increases 10-fold by homo-dimerization(please see file; The Fascio Effect).   The specificity of IgG seems to be 100-fold lower than pentamer IgM due to the Fascio effect.

Then, 2 (1) x 10 (2) x 100 (3) = 2000-fold difference seems to be occurred between HPLC-photometric method and ELISA method.

Therefore, ELISA is not quantitative and not sensitive; i.e., it gives only false value at ng/mL.

Further, I must sadly repeat about the differences in determination value of biotin between HPLC-photometric method and binding assay method; i.e., such a biotin assay as

"Anal Biochem. 1986 Apr;154(1):367-70.

A sensitive enzyme assay for biotin, avidin, and streptavidin

Edward A. Bayer, Haya Ben-Hur, Meir Wilchek

Department of Biophysics, The Weizmann Institute of Science, Rehovot 76100, Israel

Reciprocal enzyme assays are described for the vitamin biotin and for the biotin-binding proteins avidin and streptavidin. The assays are based on the following steps: (a) biotinylated bovine serum albumin is adsorbed onto microtiter plates; (b) streptavidin (or avidin) is bound to the biotincoated plates: (c) biotinylated enzyme (in this case alkaline phosphatase) is then interacted with the free biotin-binding sites on the immobilized protein. For biotin assay, competition between the free vitamin and the biotinylated enzyme is carried out between steps (b) and (c). The method takes advantage of the four biotin-binding sites which characterize both avidin and streptavidin. The method is extremely versatile and accurate over a concentration range exceeding three orders of magnitude. The lower limits of detection are approximately 2 pg/ml (0.2 pg/sample) for biotin and less than 100 ng/ml (10 ng/sample) for either avidin or streptavidin."

gives only a false value with detection limit as 0.2 pg of biotin.   This detection limit is not true since linear calibration line or linear regression line or linear correlation line is no be obtainable by the assay using binding-protein.

True HPLC-biotin determination method is presented in the file (Netherlands biotin).   HPLC-photometric method gives a true value with detection limit 17 pg of biotin.