A conservation map plots how similar a stretch of DNA is between species. The peaks are tall where the sequence is almost identical across species; valleys are where sequences differ. The idea: if a sequence is conserved across many lineages, it's probably doing something important — selection has held it in place. If a sequence is variable, it might be junk, or it might be doing something lineage-specific. This is a first-pass tool, not a verdict. Conservation suggests function, but absence of conservation doesn't prove uselessness. It's a place to look further, not a final answer.
— 336_lec14_06
A — Two kinds of change in a gene nothing is acting on
The more important a gene is, the less variation you'll see at it — not because mutations don't happen there, but because mutations that do happen there never persist. The things you see in the world are the outcome of having not died.
— 202_lec16_07
Predict first. A stretch of DNA that nothing is acting on. Count the changes that alter the protein against the changes that do not. The ratio will sit:
Changes arrive at the same rate everywhere. Some alter the protein and some do not. Set what it costs to alter the protein, and count both kinds after they have been sifted.
changes that altered the protein: —
changes that did not: —
the ratio: —
changes that did not: —
the ratio: —
A negative cost means altering the protein is favoured rather than penalised.
B — Make one kind of change costly
Negative selection — genes that don't vary, vary the most. The more important a gene is, the less variation you see in it, not because mutation doesn't happen but because the mutations that do happen don't persist.
— 202_lec16_07
TODO: selection-on-codons sim. s slider; dN/dS readout; the dN/dS vs s curve. SPEC: Make non-synonymous changes weakly deleterious. Watch dN/dS fall toward 0. Now make non-synonymous changes strongly favored. Watch it rise above 1.
C — A sliding window
Two kinds of "same gene." An ortholog is the same gene in different species. A paralog is the same gene duplicated within one individual. They came from a duplication event. The orthologs are usually more similar to each other than the paralogs are. That's a clue that the duplication is older than the species split, and that paralogs have been free to drift apart since.
— 336_lec11_05
TODO: sliding-window dN/dS analysis. Show conserved core + variable surface for an immune gene; site-class posterior. SPEC: Compute dN/dS along the gene, not over the whole sequence. Some sites are conserved (dN/dS < 1); some are evolving rapidly (dN/dS > 1). The signature of selection isn't uniform — it lives in specific residues.
D — Five genes, five patterns
The action potential is just "I have become excited." That's all a neuron can do — like muscles only pull, neurons only fire. There's no extra information in the signal. The brain has to know where the neuron came from to interpret it. The location of the neuron carries the information; the firing just says it happened. *(By analogy: dN/dS tells you "selection happened." Which kind requires context the number itself does not carry.)*
— 230_lec12_abdomenSpine_03
TODO: five-gene classification. Stretch: connect to S18 scaffold. Non-trivial code mod: take a 6th gene of the student's choice and classify it. SPEC: Classify five genes by their dN/dS pattern (scaffold S18 already drills this): MHC, Histone H3, GULO pseudogene, SARS-CoV-2 spike, mitochondrial cytochrome b. Predict each before you see the answer.