I think the logarithm regression with no weighting ends up producing results that are similar to the linear with one over x-squared weighting because of the way that the logarithm reduces the differences between large numbers.

The reason that Skyline has the logarithm option is that some labs required it as part of their procedures.

We typically recommend that you only use the calibration curve to figure out what the linear range of your analyte is. Once you have determined that your analytes are in the linear range, what we recommend is that you include a single external standard on every plate that you run and calibrate against that single point calibrator using linear through zero regression. If you have more than one plate's data in your Skyline document you can set the "Batch Name" on each replicate and external standard so that each plate is calibrated against it's own external standard.
--Nick

Thank you very much for your return Nick!

We are analyzing a curve calibration of one peptide that apparently has a better performance (higher R value and positive value LOD) when applying quadratic method than linear in Skyline, once applying this method, we observed a negative LOD value. Do you have any recommendation to decide the best fitting for a peptide?

We are using heavy peptides (SIS) for the calibration curve since our matrix of interest contains the endogenous counterpart. We are going to use this calibration curve for the absolute quantification of the endogenous peptide. So we are wondering if the external standard you are commenting about would be the heavy peptide spiked in the experimental sample (unknown concentration)?

Thank you very much in advance. Best. Daniela