Models for Inverted Duplication Formation
"Flying Walendas"

This is a single-step model
Everything is done at once
Suggested by some ideas of Sue Lovett
A pair of short inverted repeats snap back and anneal
DNA extends to second such pair, strand switches and anneals
New material can be redrawn as hanging down
Replication fork is restored
We reject this model, as it requires 2 improbable, nearly concurrent replication-dependent events.

"Slytherin"

This is a multi-step model involving sickening intermediates (perfect palindromes).
Snap-back at quasi-palindrome #1.
There are many in the vicinity of lac, on both sides.
Strand-switch at quasi-palindromes #2.
Generate large sickening palindromes.
Extruded cruciforms or lollipops act as replication blocks.
They are intolerable over the long haul.
They hang around long enough to allow some amplification and metabolism.
Remodelled by asymmetric deletions at repeats, forming so-called "short joins" (SJ's).
Speculation: Could this be a common mechanism for most selected amplifications?
We love this model!
It requires only deletions, no recombination or long-distance pairing.
It allows time to do things.
It uses the same tools as linear array remodelling.
See also mechanisms for generating TIDs from IS3 elements in F'128.