A common complaint about ML is it doesn't get to mechanism, just correlation. But in this case, we can mutate these amino acids and test if that affects lambda-max. In many cases (like 181 and 261 above), scientists already mutated them and they are tuning sites, so the ML is working!
