If you're capable of addint a standard, a simple approach is to add a spike (fixed amount) of a standard (similar to your analytes) to all your samples prior to analysis. Then compare analyte/std ratios among your different samples. Dividing the analyte signals by the std will make all the results from the chromatograms directly comparable. We do this when monitoring urine metabolites by LCMS. Hope this helps.

Thank you Daniel. This is what we have thought for prospective analysis but for my problem is with the data that has been already generated

I see your challenge, now. Is there anything in your runs that you might reasonably believe is constant, maybe even an contaminant, or a reagent that was used in the prep? If so you might try that as your std. Another approach that is not as ideal as an internal standard, but can still help is to use S/N rather than peak areas or peak heights. Peak areas can increase or decrease because of the variations between runs, regardless of the amount of sample. The S/N, however, tracks more honestly with concentration, because both signal and noise tend to increase and decrease comparably with run variations. It's not perfect but it may allow you to get some value from the data you already have. Best of luck!

Thank you so much this will help a lot

Since samples are gone. You can construct a standard curve subjected to the same matrix and extraction procedure and use that. This approach will also give you processing (extraction, matrix effect etch) standard deviation that will allow you to determine limit of detection of the procedure.

Hello Gerardo, your suggestion sound interesting. May I please request you to explain me this in detail.

Thank you.

Alpesh

You may get some idea about what's happening in your experiments through some of these suggestions, e.g., Daniel King's, but - without having internal standards in your original samples - you can never be sure of any conclusions you try to draw based on the numbers, and it will be difficult to publish anything.

John, I agree that without the internal standard the results he got may be wasted and will never be published depending on the reviewers pre conceived biased towards certain approach. I apologize for not realizing that that publication is the main reason for the question as I have been in Industry for too long.

The options are:

1) Throw away the data from presumably irreplaceable samples or

2) Try to make use of it

I voted for the latter assuming the samples are irreplaceable.

The approach I am proposing is to use statistical evaluation of the path it took to obtain the results and provide certainty of the estimate. We do not know the coefficient of variation (this varies according to the level of analyte being measured) if a known amount of sample is subjected to the same process over and over again and how many replicate samples we be needed to test for the purpose of comparison between 2 values from different samples. Once standard deviation is known we can do power analysis to tell us whether Alpesh analyzed enough number of samples to be able to conclude with sufficient power and probability the differences between his samples are significant. In the end, Alpesh may not have enough "n" to conclude any differences between his samples. If publication needs to be a positive results this will not qualify.