Well first I am not in the field of Biology or Medical Sciences. Since these days we are waiting scientists to tell us if the Indian variant of SARS Cov 2 is more transmissible than the original virus, so these two questions come to my mind:
Do scientists confirm increase in transmissibility based on empirical data?
Can scientists know how transmissible the variant is based on how well the spike protein binds to receptors of the cell, for example by doing a simulation of the 3D folded shape of the spike based on its code in the genome?
If you look at the features they say they look for in variants, the one they mention first is "spread more quickly" - the only way to tell that is to look at the genetics and how often the variant is turning up in a relative to a previous time-point. Bullet point 2 can only be seen by testing - do people who have had an infection, get infected again? Bullet point 3 - look at the severity of disease in patients.
I am sure by now that you get the drift; surveillance is the answer - all empirical data.
(2) may or may not help, all that does is tell you the potential affinity for a receptor, which is only one part of the transmissibility. In fact, we generally need to wait for the emergence of a variant (no point chasing each one; every infection will have some form of variant in some of the virions produced - they exist as a quasispecies), before modelling will take place. In addition, modelling is actually a pretty poor method of working out whether the virus will bind to a receptor. Check out this paper and see that many of the Mustelid family have poor binding by modelling, but in reality they actually get infected easily by SARS-CoV-2.
Many other factors come into viral transmission, such as viral titre in the patient, number of viral particles being shed in each droplet/aerosol, survivability of the particles in the air/droplet/aerosol etc.
Here's a nice paper from PNAS that covers the biological features of transmissibility based on a statistical/modeling analysis. While this doesn't specifically address SARS-CoV-2, it does cover a bunch of the features that are looked at for increased transmission in viruses overall.