Editing Openai/69177fbb-14c4-800f-bc57-ccb5a1d88c95 (section)

=== Assistant: Nice model — you’ve captured the key pieces and the right sorts of uncertainty. ===
Nice model — you’ve captured the key pieces and the right sorts of uncertainty. I’ll restate your points as I understand them, flag what’s well-grounded vs speculative, correct a couple of technical bits, and offer a tightened model you can use going forward.

===== 1. Many genes implicated in ASD are older than the sapiens–Neanderthal split. =====
# Both species plausibly had individuals showing autistic-style phenotypes, but the proportion in Neanderthals is unknown.
# Most Neanderthal cognition-related alleles were purged from the modern human genome after admixture because of hybrid incompatibilities.
# In Homo sapiens, ASD-linked variation persisted because component traits were often adaptive in moderation; extreme combinations were pruned by selection.
# You think there may have been positive selection on cognitive traits (including some alleles linked to ASD) during the Holocene, with a possible plateau or reversal around the industrial era / late 1800s.

===== - Polygenic, old components. Many biological pathways implicated in neurodevelopment are evolutionarily ancient; some alleles affecting neurodevelopment existed before the split with Neanderthals. =====
* We can’t infer phenotype from bones. Behavioral/psychological inference from archaeological remains is extremely indirect.
* Introgressed Neanderthal DNA was nonrandomly distributed. Studies show depletion of Neanderthal segments in regions of high brain expression and in testis/reproductive genes — consistent with selection against hybrid incompatibility.
* Mutation–selection balance model for ASD. That framework — useful midrange traits + rare/strong deleterious combinations producing clinical ASD — is a standard, evidence-backed explanation for persistence of risk.

===== - On “ASD genes are older than the split”: many genes and pathways involved in neurodevelopment are ancient, but specific risk variants identified in modern GWAS can be recent. Also, ASD risk has two partly distinct genetic contributors: (a) common, small-effect polygenic variation (older and shared) and (b) rare, often de novo, large-effect mutations (recent). So “ASD genetics is older than the split” is true for the broad pathways but not universally true for every risk allele. =====
* On measuring ASD risk in Neanderthals via GWAS: you cannot run a meaningful GWAS on a handful of ancient genomes. GWAS requires large samples and the same population structure. What you can do is: (1) check whether loci implicated by modern ASD GWAS overlap regions introgressed from Neanderthals, and (2) look for Neanderthal alleles at known risk sites. Those approaches test for overlap/enrichment, not “ASD prevalence in Neanderthals.”
* On Holocene positive selection for intelligence / ASD alleles: that’s plausible as a hypothesis (selection for cognitive/learning traits as societies grew) but it’s controversial and tricky to demonstrate. Human GWAS signals for educational attainment and cognitive proxy traits show some signatures consistent with selection, but these results are confounded by population structure, stratification, changing environment, and demography. Claims of directional selection on “intelligence” during the Holocene need careful, conservative interpretation. The idea of a late-industrial reversal is likewise plausible (relaxation of selection due to medicine, differential fertility, etc.), but again empirical support is mixed and contested.

===== 1. Deep background: Core neurodevelopmental pathways are ancient and shared across hominins. This creates a substrate for variation in both Neanderthals and sapiens. =====
# Two genetic layers: - Polygenic background (many old alleles with small effects; these shape continuous cognitive variation). - Rare large-effect variants (often recent or de novo; these can shift an individual into the clinical ASD range).
# Admixture effect: When sapiens and Neanderthals mixed, some Neanderthal alleles persisted (mostly in skin, immunity, metabolism), while many alleles affecting brain development were purged because they produced maladaptive interactions in the sapiens developmental program (hybrid incompatibility). That purging is about genetic context, not the intrinsic “goodness” or “badness” of a phenotype in the donor population.
# Selection on variation in sapiens: Within sapiens, selection acted on cognitive and behavioral variation. Many component traits of ASD could be neutral or advantageous in certain ecological niches; extreme clinical presentations are kept rare by selection and by the stochastic nature of mutation.
# Recent millennia dynamics: There may have been modest directional shifts in allele frequencies affecting cognition during the Holocene (e.g., selection for literacy-friendly, cooperative, or technical skills), but separating true selection from demography/population stratification is hard. Modern changes (medicine, social structure, fertility differences) may have altered selective pressures, producing shifts in the distribution of clinical risk—however, the evidence is mixed and methodological noisy.

===== - “Could Neanderthal genomes be analyzed for ASD risk?” Not by GWAS on Neanderthals. You can look up whether modern ASD-associated loci are present in Neanderthal sequences or whether introgressed segments overlap ASD loci. That’s informative about overlap/enrichment, not prevalence. =====
* “Was there ASD in both species?” Very plausible at the level of individual variation (there’s no reason to imagine hominins lacked cognitive diversity). But we have no direct way to estimate prevalence in Neanderthals.
* “Did selection favor cognitive traits in the Holocene?” It’s plausible some cognitive traits were under selection during agricultural and post-agricultural transitions, but the signal is noisy and contested. Treat that as a plausible hypothesis, not an established fact.

===== - Step 1: check enrichment of modern ASD GWAS loci in introgressed Neanderthal segments (this is a standard genomic overlap test). =====
* Step 2: examine whether alleles at high-impact ASD loci are present in available high-coverage Neanderthal genomes.
* Step 3: model allele-frequency trajectories for cognitive-proxy GWAS hits across ancient DNA time transects to test Holocene selection (requires ancient human aDNA panels). I can sketch the exact analyses you’d run and the interpretation pitfalls for each.

Bottom line: your core intuitions are good. Tighten two things—(A) don’t treat every ASD risk allele as ancient, and (B) treat Holocene directional selection claims as hypotheses that need careful genomic/demographic control. Want me to draft the specific overlap/enrichment test you could run (methods, stats, and likely outcomes)?