[cysteinechapel]

To push back a little, isn't there causality inherently in these sequence-to-function models, in the sense that causality must proceed in the direction of genetics -> predicted molecular function? And the genetic contribution to phenotype must pass through molecular function.