The design inference

For example, in the common bacteria E coli, there are three genes in the B-gal system : a permease (lacY), which allows B-galactosides into the cell; a B-galactosidase (lacZ), which digests the B-galactosides into usable sugars (galactose and glucose, through allo-lactose); and an enhancer (lacI), which increases production of the other two lac proteins about 500 to 1000x in the presense of B-galactosides. This is an 'irreproducbly complex' system - remove any one of the proteins (lacY, lacZ, or lacI), and the cell can't use B-galactosides as a carbon source.

​

Back in 1982, Barry Hall removed the lacZ gene from a strain of E coli, then plated them to see how long it would take the bacteria to re-develop lacZ function. It took less than 3 weeks.

​

The bacteria had another protein (completely unrelated to lacZ) that could catalyze the needed reaction, but at about one thousandth the needed levels (even if 100% of the protein in the bacteria were this, it wouldn't be enough). A mutation increased its ability to catalyze the reaction to usable levels. A second, later mutation altered the protein again, so that the end product could activate the lac system again - IT WAS NOW REGULATED (only active in the presence of substrate); the old lacI and lacY proteins were now part of a new system (which is now, again, irreproducibly complex).