Draft skeleton. Stage scenarios and anchor quotes are in place; the simulator and code panels are not yet wired.
Do your cells know what a gene is? No. The thing that lets us separate genes is dynamics that happen later — recombination. As nice as it is to split things off into alleles and loci, in actuality what matters is the physical reality of the DNA strand. A locus is just a chunk of DNA selection hasn't recombined apart yet. Whether selection sees two genes as one locus or two separate ones depends on the recombination distance and how strong selection is.
— 202_lec19_05
A — Free recombination: genes are the unit
Simulate two loci with high recombination between them. Selection on one locus doesn't drag the other. The cov(w, z) at the gene level dominates the chromosome level. Each gene is its own selection unit.
Selection acts on traits, but you inherit nucleotide sequences. There's a massive disconnect between what gets inherited and what gets selected.
— 202_lec17_01
TODO: two-locus sim. Recombination-rate slider; show selection sweep at locus 1 leaves locus 2 alone when r is high.
B — Tight linkage: the chromosome is the unit
Drop the recombination rate. Now selection at one locus drags neighbors. The cov(W_chr, Z_chr) term grows; the within-chromosome cov term shrinks. The chromosome behaves as one unit because selection can't pull its components apart fast enough.
Can I do any recombination in an inverted region if the partner isn't inverted in the same area? No. If there's a bunch of genes in that area, can I recombine them at all? Nope. They're stuck together. Inversions make super-loci — a whole bunch of things you can't recombine until and unless the inverted form becomes common.
— 202_lec12_04
TODO: tight-linkage sim. Selective sweep at locus 1; track diversity at locus 2 as a function of physical distance.
C — The diagnostic: how big is the sweep?
Run the Price ratio at the chromosome scale. The fraction of variance that lives between chromosomes vs within chromosomes is the diagnostic. Pair this with the empirical signature: a sweep's footprint is the region over which selection has been faster than recombination could erase it.
Selection is self-defeating. Every time I select, I reduce my additive genetic variation. Selection is fast — way faster than mutation, faster than drift in almost every circumstance, sometimes faster than recombination. When that happens, selection can drag bad alleles to fixation if they're physically linked to good ones. Selection on phenotype is blind to the linkage.
— 202_lec19_03
TODO: sweep-footprint analyzer. Width of zero-diversity region as a function of s and ρ; compare to Price ratio at the chromosome level.
D — The dachshund's leg
Real data: the 20-megabase region of zero diversity around FGFR3 in dachshunds. Estimate the strength of selection from the sweep width. Then apply the same diagnostic to a wild population (e.g., a Florida Scrub Jay locus under inferred recent positive selection).
FGFR3 is the receptor for fibroblast growth factor — the protein that tells limb cells to grow. In dachshunds, we bred them to have short legs — under positive selection. If you look at dachshund DNA versus non-dachshund DNA around FGFR3, dachshunds have basically zero genetic diversity in a region of about 20 megabases on either side. That's a really big region of zero diversity. The selective sweep tells you how strong selection was: a wider area of low variation means selection had to outpace 20 million base pairs' worth of recombination.
— 202_lec19_04
TODO: dachshund FGFR3 case study. Estimate s from sweep width; .R export. Non-trivial code mod: rerun the analysis on an inversion (e.g., the 16p11.2 human inversion) and ask whether the whole inverted region is the unit.