This can help you: https://www.mathworks.com/help/stats/pca.html

I tried this but the loading plot i am getting by the way they are suggesting, is not the kind which are being shown in most of the publications in case of raman spectroscopy. Can u please tell if loading plot changes according to the data? And how to find what should be done for which data?

https://www.mathworks.com/help/stats/examples/partial-least-squares-regression-and-principal-components-regression.html

The last picture is about loadings

It is about processing of spectral data.

If it is not good for you, please show picture of what you want, I'll try to find how to obtain the result

Dear Zlatin

The loading plot, of course, changes depending on the data. Your spectrum may have some variances that the ones in the literature do not have. You can look for which variables have the most variances (which you do not want it) and exclude some part of the spectra. There are many feature selection algorithms that you can apply on your spectra in the literature. By the way the number variables seems strange to me, did you already throw some part of the spectrum or do you really have 450 wavenumbers in your spectrum?