Really interesting article. By the end, they admit that what makes Delta worse is the sum of the parts, and not just a single mutation in the spike protein.

This is a general property of biology: it's almost always the sum of the parts.

It reminds me the trajectory of computer sciences, where a lot of our (conventional) systems are carefully coordinated part orchestras. I.e., an engineer knows all the classes/functions/modules and how they interact. Compare with our recent multi-layer neural networks, where nobody can say what the coefficient at index [22, 2, 432, 91, 42] does exactly. I wonder if, similarly, bio-research will be able to move more towards automatic discovery of desirable systems.