I am working with metabolite data from an experiment that looked at the exometabolome in filtrate of bacteria grown in a medium made with natural seawater. Because natural seawater is a complex mixture, the data also includes analysis of a sample of the medium as a "blank". I was planning on subtracting the value from this medium blank for each metabolite identified in the filtrate to correct for background presence of that metabolite in the medium, but many of the values for metabolites in the medium are below the specified limit of detection. Should I subtract the blank anyway even if its value is below the LOD?
If "values for metabolites in the medium are below the specified limit of detection",
metabolites are analytically speaking zero, e.g. Article Use of cut‐off values on the limits of quantification report...
Dear Gretc
I would say it depends on the purpose of the data. As such a general definition however, I do not agree with Roberto's statement that
I agree that the answer is NO. But, if I read the question correctly, you only ran one Blank. If that is the case, I would run at least 4 more on new medium mixtures prepared from well - mixed reagents and see if the blanks are constant. If not, I would be concerned if the blanks are properly characterized for your preparations.
Hi there
I hope I could help you in solving this problem. First, I should know the nature of measured signal and applied instrument. Second, the analyte signal in a real sample is obtained by subtracting measured total signal of the sample from sample blank signal. However, if the blank signal value is less than sample signal (near zero or
To subtract the blank value does not improve the accuracy of the analytical result – it only increases its uncertainty.
Let's do a Gedankenexperiment:
I analyze samples from two experimental runs. One had low reagent contamination and the control sample yields 4, my test sample 6 and my LOQ is 5.
Not considering the blank yields 6.
The other the same calibration statistics and but higher contamination and the control yields 6, the test sample 8.
Now I can subtract the control and get 2.
Of course the uncertainty is substantially higher but not considering the blank in the first case will additionally bias my result by 3x.
As mentioned: I don't advocate for blank subtraction as a general rule but I need to compare a series of experiments using the same analytical method it may reduce censoring of low values.
Good point Guilherme Luiz Scheffler . At least the matrix effect has to be evaluated.
- Badges
- Science topic
- Similar topics
- Correction