Jump to content

  • Log in with Facebook Log in with Twitter Log in with Windows Live Log In with Google      Sign In   
  • Create Account

Submit your paper to J Biol Methods today!
Photo
- - - - -

Problem with ELISA false positives and noise


  • Please log in to reply
15 replies to this topic

#1 AndreaELISA

AndreaELISA

    member

  • Active Members
  • Pip
  • 8 posts
0
Neutral

Posted 07 October 2009 - 02:25 AM

Help please kind experts.
We have an ELISA with the following protocol for a 30kD protein in unconcentrated urine.



Two monoclonal mouse antibodies were raised using the synthetically produced C-terminal 100 amino acids (Biosynthesis Inc, USA) as an antigen (Antibody Production Services Ltd, UK). One of these, APS1, was conjugated to Alkaline Phosphatase using the Lightning Link Alkaline Phosphatase Conjugation kit (Innova Biosciences, UK), whilst the other, APS2, was conjugated to biotin using the Lightning Link Biotin Conjugation kit (Innova Biosciences, UK). APS2-biotin was captured onto a 96-well streptavidin-coated plate (Nunc 436014, USA) at a concentration of 4μg /ml.

After washing, 100μl of urine or a dilution of the fragment in
buffer was incubated in each well for 1 hour at room temperature.

The plate was then washed 8 times in buffer and the secondary detection antibody APS1-Alkaline phosphatase was added to each well at a concentration of 4μg/ml (1 hour at room temperature).

After a final wash step a colourmetric agent pNPP (Sigma, USA) was added and the absorption of light at 405nm was measured after one hour. The dilution series was used to generate a standard curve by which the concentration of EN2 in each sample was measured.

To generate a decent looking standard curve we are having to make 5 measurements at each point and then take the median measurement.
We are also having to use 5 wells per sample and use a median to obtain a sensible output as individual wells vary by as much as a factor of 2.http://www.protocol-online.org/forums/style_emoticons/default/sad.gif

What are the most likely sources of our woes? All help gladly accepted

The protein we are measuring is sticky but that doesn't explain highish readings at 0 concentration.

Other ELISAs work fine on the equipment.

#2 sgt4boston

sgt4boston

    Veteran

  • Active Members
  • PipPipPipPipPipPipPipPipPipPip
  • 276 posts
1
Neutral

Posted 07 October 2009 - 11:59 AM

Hello,
Since other tests on same equipment work fine...ie you do your other curves in singleton or duplicate and this assay requires 5 (!!!) replicates...lets take it one step at a time.

You indicated analyte is "sticky" and you are using unconcentrated urine as matrix.

1. Examine the urine matrix
Can you do another test but run your analyte in plain PBS with 0.1% BSA. Run in several replicates to see if %CV has improved. Urine unconcentrated will precipitate proteins etc and pH, conductivity can be dramatically different and change over time.

2. Your sticky analyte
Is there surfactant in your wash solution? Can you let the plate/wells sit for 15-30 seconds before decant/aspiration?

These should be easy to do first.

#3 AndreaELISA

AndreaELISA

    member

  • Active Members
  • Pip
  • 8 posts
0
Neutral

Posted 08 October 2009 - 06:06 AM

Thanks very much sgt4boston

The urine doesn't seem to be the problem

our average %CV with urines is 25 (we do buffer the urines to standardise pH)
the %CV of the analyte in buffer only for the standard curve is around 28%.

the sticky analyte may also not be much of a problem as the %CV for the 0 analytes is even higher.

Hello,
Since other tests on same equipment work fine...ie you do your other curves in singleton or duplicate and this assay requires 5 (!!!) replicates...lets take it one step at a time.

You indicated analyte is "sticky" and you are using unconcentrated urine as matrix.

1. Examine the urine matrix
Can you do another test but run your analyte in plain PBS with 0.1% BSA. Run in several replicates to see if %CV has improved. Urine unconcentrated will precipitate proteins etc and pH, conductivity can be dramatically different and change over time.

2. Your sticky analyte
Is there surfactant in your wash solution? Can you let the plate/wells sit for 15-30 seconds before decant/aspiration?

These should be easy to do first.


Thanks very much sgt4boston

The urine doesn't seem to be the problem

our average %CV with urines is 25 (we do buffer the urines to standardise pH)
the %CV of the analyte in buffer only for the standard curve is around 28%.

the sticky analyte may also not be much of a problem as the %CV for the 0 analytes is even higher.

#4 Gerard

Gerard

    Veteran

  • Active Members
  • PipPipPipPipPipPipPipPipPipPip
  • 124 posts
0
Neutral

Posted 08 October 2009 - 10:18 AM

our average %CV with urines is 25 (we do buffer the urines to standardise pH)
the %CV of the analyte in buffer only for the standard curve is around 28%.

the sticky analyte may also not be much of a problem as the %CV for the 0 analytes is even higher.


%CV is not always a good indicator
result 0.1 and 0.2 has a CV of 47% but if the range goes from 0 to 100 it is very good.
If you have a calibration line the on the same OD you wil find different CV's dependent of you substract the zere standard before or after CV calculation.
Ockham's razor
Pluralitas non est ponenda sine necessitate
-- "You must assume no plural without necessity".

#5 AndreaELISA

AndreaELISA

    member

  • Active Members
  • Pip
  • 8 posts
0
Neutral

Posted 08 October 2009 - 11:14 AM

our average %CV with urines is 25 (we do buffer the urines to standardise pH)
the %CV of the analyte in buffer only for the standard curve is around 28%.

the sticky analyte may also not be much of a problem as the %CV for the 0 analytes is even higher.


%CV is not always a good indicator
result 0.1 and 0.2 has a CV of 47% but if the range goes from 0 to 100 it is very good.
If you have a calibration line the on the same OD you wil find different CV's dependent of you substract the zere standard before or after CV calculation.


Thanks, unfortunately the figures I quote are before we subtract the zero standard - so they only get worse and the range is not 0-100.

We really need to reduce the variation and in particular to eliminate seemingly random high spikes. (we rarely if ever get random low values).

#6 Gerard

Gerard

    Veteran

  • Active Members
  • PipPipPipPipPipPipPipPipPipPip
  • 124 posts
0
Neutral

Posted 09 October 2009 - 12:01 AM

Maybe it will help if you change your dilution buffer for the antibodies or to increase the protein content (BSA/HSA) in the buffer.
Ockham's razor
Pluralitas non est ponenda sine necessitate
-- "You must assume no plural without necessity".

#7 sgt4boston

sgt4boston

    Veteran

  • Active Members
  • PipPipPipPipPipPipPipPipPipPip
  • 276 posts
1
Neutral

Posted 09 October 2009 - 05:01 AM

As you indicated the %CV is high with both urine matrix and buffer. On average in either matrix you see 25%. Normally %CV by OD should be <10%.

Gerard indicated that the high CV could be due to the area of the curve you are examining. Is this high CV throughout the "linear" range of the curve? Normally, %CV is highest at the very low and very high concentrations. (if plotted it looks like a "U":::: %CV v. Concentration. You gave us average %CV.

Are you coating your wells high enough so your reaction is not at the level where the coating ends?

There should be proteins in your buffers especially your conjugate.

I am curious...you should have some unconjugate primary ab that you could directly absorb to wells; block your plate and perform the same test. This would eliminate any problems with the SA coated plate and the biotin conj antibody. Could you do this and let us know if this works?

#8 AndreaELISA

AndreaELISA

    member

  • Active Members
  • Pip
  • 8 posts
0
Neutral

Posted 09 October 2009 - 07:32 AM

As you indicated the %CV is high with both urine matrix and buffer. On average in either matrix you see 25%. Normally %CV by OD should be <10%.

Gerard indicated that the high CV could be due to the area of the curve you are examining. Is this high CV throughout the "linear" range of the curve? Normally, %CV is highest at the very low and very high concentrations. (if plotted it looks like a "U":::: %CV v. Concentration. You gave us average %CV.

Are you coating your wells high enough so your reaction is not at the level where the coating ends?

There should be proteins in your buffers especially your conjugate.

I am curious...you should have some unconjugate primary ab that you could directly absorb to wells; block your plate and perform the same test. This would eliminate any problems with the SA coated plate and the biotin conj antibody. Could you do this and let us know if this works?


Sgt4boston, Gerard. thanks.

I have recently realised that maybe an early answer I gave was wrong: the equipment does work OK with other ELISA's but I had not taken into consideration that we were not looking for the same level of sensitivity (we're aiming for 1ng/ml). We're going to try running the test on another piece of equipment and see if that sorts it out (at least to some extent).

I'll also check the level of the wells.

The %CV is not U shaped or linear, it's all over the place. I'd almost say that it's a random error, but bigger with urine than with buffer only.

We are trying to measure in the 0 to 100 g/ml range (with the key objective being to distinguish between 0 and anything else).

I'll let you know how we get on.

#9 sgt4boston

sgt4boston

    Veteran

  • Active Members
  • PipPipPipPipPipPipPipPipPipPip
  • 276 posts
1
Neutral

Posted 13 October 2009 - 02:35 AM

I suspect it is not your equipment. I think you are pushing the limits of the assay. Please extend your curve into the ug and mg/ml area ...run serial 1:3 or 1:10 dilutions and let us know if the %CV improves as the concentration increases.

#10 AndreaELISA

AndreaELISA

    member

  • Active Members
  • Pip
  • 8 posts
0
Neutral

Posted 13 October 2009 - 05:28 AM

I suspect it is not your equipment. I think you are pushing the limits of the assay. Please extend your curve into the ug and mg/ml area ...run serial 1:3 or 1:10 dilutions and let us know if the %CV improves as the concentration increases.

We will do and get back to you. By equipment I also mean wider lab environment as temp control is not always great.

#11 sgt4boston

sgt4boston

    Veteran

  • Active Members
  • PipPipPipPipPipPipPipPipPipPip
  • 276 posts
1
Neutral

Posted 13 October 2009 - 08:43 AM

Temperature should not influence CV unless the temperature was different across the area of the plate...or if you were comparing one assay to another done on different days. Your intra-assay CV is high.

#12 AndreaELISA

AndreaELISA

    member

  • Active Members
  • Pip
  • 8 posts
0
Neutral

Posted 15 October 2009 - 02:52 AM

I suspect it is not your equipment. I think you are pushing the limits of the assay. Please extend your curve into the ug and mg/ml area ...run serial 1:3 or 1:10 dilutions and let us know if the %CV improves as the concentration increases.

Yes the %CV improves with increasing concentration. We have also had a recent run with lower %CV at low levels, (luck, practice, we're not sure). It will be a while before we get a chance to do the test in very controlled conditions, but I'll let you know how we go then.

We're still struggling with too high a background signal though. Anything from our protocol that could help reduce background signal?

#13 sgt4boston

sgt4boston

    Veteran

  • Active Members
  • PipPipPipPipPipPipPipPipPipPip
  • 276 posts
1
Neutral

Posted 16 October 2009 - 09:53 AM

Is there surfactant in your wash buffer? Tween -20 or Triton X 100 are usual additives.

You can allow the wash buffer to sit in wells for 10-15 seconds before you decand.

Be sure to vigorously decant and blot between washes.

After the last wash allow the plate to be inverted for a few minutes.

What is the %CV for each of the calibrator levels...including the new ug-mg/ml levels?

For example in a troponin test the CV is very high in the low pg ng/ml level and decreases to less than 10% as one gets to 5+ ng/ml levels.

#14 AndreaELISA

AndreaELISA

    member

  • Active Members
  • Pip
  • 8 posts
0
Neutral

Posted 27 October 2009 - 12:57 AM

So, our high CV% have mysterioulsy gone down to the point where triplicates now easily suffice. Which is nice for us but not so useful for the advancement of how to improve ELISA's as we don't know what is different now than before.


We're now going to try and reduce background by trying other blockers. So far we have tried standard stuff, skim milk, BSA, Ovalbumin. BSA has given the best responses so far. We're thinking of trying some of the more exotic non protein blockers or non mammalian blockers. Any recommendations either of what to try or how to try it?

Cheers

#15 sgt4boston

sgt4boston

    Veteran

  • Active Members
  • PipPipPipPipPipPipPipPipPipPip
  • 276 posts
1
Neutral

Posted 30 October 2009 - 11:03 AM

For blocking try commercial products from Surmodics and East Coast Biologicals. There is also a blocking peptide fragment from Toyobo.

I suggest you also coat plates directly with ab. It may eliminate a potential source of problem.




Home - About - Terms of Service - Privacy - Contact Us

©1999-2013 Protocol Online, All rights reserved.