Linearity Fail - Data of any value? (Apr/02/2010 )

Hi,

I've been testing out dilution factors for some samples before I run my ELISA and found the results to not be linear.
For 3 different dilution factors (serial 2x dilution), the unadjusted values were pretty much the same. The values decreased slightly as the dilution factor increased, but were far from a 50% decrease. I suppose it's probably due sample matrix interference (or heterophilic Ab interference?).

I would try a higher dilution factor to overcome this, but at the above dilutions they're at the low end of my standards so I'm not sure if greater dilution would help.
However, what I'm mainly wondering is, I was told to run the samples in the ELISA anyways, at one of the above dilutions. Is the data I obtain from this of any use/value?

I was thinking that the data won't be accurate (the ELISA results won't reflect the actual levels in the samples).
There are samples from different treatment groups, so I guess they mostly want to see a difference between groups... would we be able to (validly) determine a difference between groups? I think the interference will mute a lot of the difference between samples, so I don't think it's likely that we'd see a difference... but if there is, can the results be trusted?

Thanks in advance!

Hi Liza... why dont u run a sample control... by this atleast the matrix interference will be clear... if tat does not give any signal, then it may be due to the ELISA components right frm the diluent to antibody!!!

Also to rely on the accuracy u can do a spike at the lowest point of the curve so tat u can decide wether to trust or not the values obtained... and if its a qualitative observation as comparision between the groups and everything else is same in the two groups... and if u are sure about the accuracy and precision... u might be able to use the data... or else it will have to be trouble-shooted ion which case we might need more details!!!

Alternativey if u sure that it is the matrix of the sample tat is interfering.. u can buffer exchange it and then analyse!! (like a centricon where u can buffer exchange as well as concentrate using the lowest cut off filter so tat ur above dilution problem can also be sorted out!!)

Hope it helps a bit..

Best Luck

A few questions comments...
1. Is your elisa commercial or home grown? Commercial...you can get info from manufacturer.
2. What is the size MW of the analyte...is this a competative test drug/steriod, etc...small MW will pass thru the pores of the filters...(FYI heterophile antibodies will remain with low MWCO)
3. What is the diluent? The diluent should be the matrix of the samples without the analyte present and should match the matrix of the standards/calibrators.
4. You indicated you have an accuracy and not a precision problem (if you run replicates is your %CV < 10% across the linear range of your curve?)