This depends a bit on what you are trying to do. In most cases, you will add known levels of amino acids to samples of plasma and, if appropriate internal standard or standards. To account for amino acids present, you will need to determine where the response (or relative response to the internal standard) for the amino acids (typically plotted on the Y axis) would be zero. This will be the point at which the the plot (normally a linar plot) crosses the X axis. The value on the X axis would show you how much you need to add all concentrations to move the curve through the intercept. (The vaule read from the X axis is the negative value of the background concentration.) You need to be careful with this calculation it makes a couple of important assumptions: 1) the curve is linear below the lowest level you can measure. 2) the resonse is only due to the analyte of interest.
If your response in the range over which you can acquire data points (don't forget to run a sample into which no analyte was added) is not liner - you have a problem already.
You have to be very careful on how you detect your analytes. In some cases, you may be able to report your analytes only as "not greater than" some the value you determine by reading the result from the calibration - because you can not prove that you do not have an interferece.
