Rohail, your question leads to even more questions for me..

- How do you know which samples are true pos and true neg? For cut-off ranges you need this information.

- Do you have the antibodies used in the ELISA? You could use them to deplete some samples from the antigen and use those samples as your true negative.

It's not ideal, but now that you have measured some samples you may use some of those samples to make your standard curve. Pool some (10?) of the highest-level samples (call this your 100% pos mix) and some of the lowest level samples and call this sample (could also use the depleted sample as your negative/low-level mix, as suggested above) and call this mix 0%.

Then, make serial dilutions of these mixes in increments of 10% or so. Use these diluted mixes as you standard curves. This will give you a rough estimation of the amount of signal your ELISA generates in low-to-high antigen samples and may provide some clarity for your experiments 

Also, you could use one medium-low and one medium-high sample to include in every plate; this sample should give the same percentage of your antigen in every run. Variations in these samples gives you a feeling of the inter-plate variability.

I agree with Wesley's ideas but need more information.  Firstly, it is much easier to give help with more details about the assay, antigen, coating conditions, wash buffers , sample being measured & dilution used, conjugate etc.  Secondly,  What readings are you getting for your true positives, true negatives and what are your blank readings.

Thank you for your helpful answer.actually i used several healthy controls with the patients samples to test my antigen.those samples have significantly positive compared with healthy controls,i called them as positive samples but some samples close to healthy control or not significantly positive/negative,i called them false positive/negative.

I coat 200ng/100µl/well antigen in PBS overnight,wash 4 times with PBST and block with 1%BSA in PBS at RT for 2hours,wash 4 times with PBST and apply serum 1:2000diluted at RT for 2hours,wash 6 times and add 1:50000 diluted secondary antibody conjugate with enzyme at RT for 2hours,wash 8 times and apply TMB substrate and incubate 20 min at RT and read it at 450nm

Blanks values ranges in between 0.05 to 0.07

Healthy control ranges in between 0.3 to 0.6

False positive negative samples ranges in between 0.5 to 0.9

True positive samples ranges in between 0.9 to 1.3

Thank you

Your blanks are good and will gain a small advantage by reading at dual wavelength and suggest that you try reading at 450 - (630 to 650) if you can.  This may help to lower the readings of the healthy control, false & true positive and give you a more significant result..  I would choose the highest reading positive samples you have and create a dilution curve, probably starting at 1:125 and titred with doubling dilutions until near  blank values, 1:64000 or beyond.  This titre should also  include some false positive, Healthy control and blank values and preferably across one plate for consistency.   A one in 1:2000 dilution is a good starting place but with your assay, you may need to working at a different dilution. I would like to see the graph of these results.  

I query if the 1:50, 000 dilution of the conjugate has been chosen by a dilution series as a working strength.  Your blanks are good but is this dilution of your conjugate restricting the values of your positive results?

Thanks sir and i will follow your directions.i already optimize assay on different serum and secondary AB dilutions and this was the best condition that i explained you earlier.you can find the attached file of my ELISA data.Thanks

As the other colleagues said before, you need a training population to optimize your ELISA. The (additional) information about the real (positivity) or real negativity has to come either from another assay or from other diagnostic procedures or assays or from the clinical symptoms or the overall clinical picture.

To optimize an assay, you need to look for the best positive/negative (P/N) ratio. Looking for your assay procedure, the coating concentration with 2 µg/ml seems to be OK. Mostly coating at basic pH works much better than at neutral pH. Therefor a 0.1 mol/L carbonate buffer pH 9.6 is often used. Incubation overnight at 4 °C is OK. Now remove the coating volume and store the plates frozen until use. Blocking is always a kind of voodoo and not helpful (only useful in commercial assays wich needs to have a long term stability at room temperature). Blocking is done with the first washing with a detergent containing buffer (e.g. PBS 0.1% Tween 20).

The next important part is the incubation buffer. Normally PBST is used. 1 … 3% BSA as an inert protein is fine. In many cases we used a double NaCl-concentration = 0.01 mol/L phosphate buffer pH 7.2, 0.3 mol/L NaCl, 0.1% tween, 1% BSA (or I replaced is by a hydrolyzed gelatin available as drug used as plasmaexpander). All this components, the detergent, the inert protein and especially the NaCl concentration (between 0.15 … 0.45 mol/L) have a huge impact on the P/N-ratio. The second important part here is the sample dilution. Also this needs to get optimized for the P/N-ratio.

All the other steps are less important, conjugated antibody, substrate reaction, reaction times etc.

As positive controls (or standards) often specific monoclonals are helpful. You can detect the human and the murine immunoglobulins by Protein-A/G-HRP.

When you have finalized this, you need a second training population. Here you can apply the first time a fourfold table getting the numbers for right negative, right positive, false negative and false positive. This allows the calculation of the specificity and the sensitivity. The Youden index (Y=Se+Sp-1) will summarize both within one value. By changing now the assay conditions again (see buffer) you can achieve an optimum as well.(look for the figures in the attached paper)

With this trained assay you can go into the real population to evaluate your test.