I am not sure, but I think what most scientists end up doing is using Linear regression and 1/(x*x) weighting.

The reason that you need to use 1/(x*x) weighting is that otherwise the higher concentration points end up controlling the path of the calibration curve because of the way that least squares linear regression ends up being dominated by the concentration levels with higher variance.

However, with the 1/(x*x) weighting you then run into the problem that the calibration curve ends up being controlled by the points at the very low end of the concentration scale, and, then, the calibration curve does not seem to pass anywhere near the points at the high end because they are weighted so little. This is happening because there are points are the low end that are really below the limit of quantification, and the positions of these points are changing the trajectory of the calibration curve. What you then need to do is right-click on the lowest concentration points and choose "Exclude from calibration".

In the future, we hope to implement a new way of calculating the LOD and LOQ that involves the bilinear fit, but it will probably be another year before that makes it into Skyline as a real feature.
There is an option to do a "bilinear" regression on the quantification tab, but I think we recommend the straight line regression and manually excluding concentration levels.

We recommend "2" for the MS Level.
20% is a good values for Max LOQ bias and Max LOQ CV.
Some people calculate their LOD by doing the blank plus 2 standard deviations, and some use 3 standard deviations. I don't know which is more common.

Note that once you have figured out what the linear range of your compounds are, and you actually want to quantify experimental data, we recommend using a single point calibrator as outlined in this paper:
https://pubs.acs.org/doi/abs/10.1021/acs.analchem.8b04581
-- Nick

Thank you Nick!

I did try the Bilinear and the linear using Log transformed X and Y axis with weighing of 1/x*x. I did see people have preferred 1/x*x for regression weighing.
I have re attached my files with updated curves.

I think, my curve looks pretty good. I did get a LOD and LOQ, however, I will be spiking double the amount of the LOD and LOQ.

I came across few papers such as **Mass Spectrometry-Based Assay for Targeting Fifty-Two Proteins of Brain Origin in Cerebrospinal Fluid** where authors have used No weighing.
And **A study protocol for quantitative targeted absolute proteomics (QTAP) by LC-MS/MS: application for inter-strain differences in protein expression levels of transporters, receptors, claudin-5, and marker proteins at the blood–brain barrier in ddY, FVB, and C57BL/6J mice** used 1/x as weighing