Embryo Genetic Selection

With 1 in 36 babies being born through in vitro fertilization every year, IVF has been firmly established as reproductive healthcare 1. Each potential baby is screened for gross genetic disease but otherwise the embryo to implant is usually arbitrary.

Modern computational biology utilizes polygenic risk scores (PRSs) to evaluate quantitative traits controlled by thousands of intersecting genetic variants. With sequencing costs falling from hundreds of thousands of dollars to less than a thousand, embryo selection based on other factors (such as IQ) is likely to get cheaper 2.

The policy debate surrounding polygenic embryo selection frequently centers on proposals for state-enforced prohibition or strict diagnostic limits. Though embryo selection is likely to be superseded by more effective methods such as genome synthesis [3] and the current technology's low cost-benefit, future restrictions would create a regulatory premium that isolates the technology for the wealthy.

Wealthy patients routinely travel abroad to access treatments unavailable in their home countries, a phenomenon sometimes called "medical tourism" where the middle class is unable. [expand with more evidence]

Genetic engineering would allow an artificial transfer and a build-up of advantage beyond current methods which are comparatively ineffective: assortative mating, shared environment effects. Here’s a toy example: assuming 50% variance in intelligence due to IQ over 5 generations of 8 viable embryos per generation we can simulate the effects of genetic selection. The results indicate that genetic selection over 5 generations would result in a gap of more than 1 standard deviation: a significant gap which would allow for the elimination of regression to the mean.

import numpy as np
import pandas as pd

def simulate_multigenerational_selection_fixed(
    generations=5,
    embryos_per_generation=6,
    heritability=0.5,
    trait_mean=100,
    trait_sd=15,
    iterations=10000,
    r_assortative=0.35,
):
    final_control_scores = []
    final_enhanced_scores = []
    final_assortative_scores = []

    genetic_sd = trait_sd * np.sqrt(heritability)
    environmental_sd = trait_sd * np.sqrt(1 - heritability)
    recombination_sd = genetic_sd / np.sqrt(2)

    for _ in range(iterations):
        # CONTROL
        gen_control = 0.0
        for _ in range(generations):
            partner = np.random.normal(0, genetic_sd)
            mid_parent = (gen_control + partner) / 2
            pool = np.random.normal(mid_parent, recombination_sd, embryos_per_generation)
            gen_control = pool[0]
        final_control_scores.append(trait_mean + gen_control + np.random.normal(0, environmental_sd))

        # ENHANCED
        gen_enhanced = 0.0
        for _ in range(generations):
            partner = np.random.normal(0, genetic_sd)
            mid_parent = (gen_enhanced + partner) / 2
            pool = np.random.normal(mid_parent, recombination_sd, embryos_per_generation)
            gen_enhanced = np.max(pool)
        final_enhanced_scores.append(trait_mean + gen_enhanced + np.random.normal(0, environmental_sd))

        # ASSORTATIVE + SELECTION
        gen_assortative = 0.0
        for _ in range(generations):
            partner_mean = r_assortative * gen_assortative
            partner_sd = genetic_sd * np.sqrt(1 - r_assortative**2)
            partner = np.random.normal(partner_mean, partner_sd)
            mid_parent = (gen_assortative + partner) / 2
            pool = np.random.normal(mid_parent, recombination_sd, embryos_per_generation)
            gen_assortative = np.max(pool)
        final_assortative_scores.append(trait_mean + gen_assortative + np.random.normal(0, environmental_sd))

    return pd.DataFrame({
        "Control (Natural)": final_control_scores,
        "Enhanced (Standard)": final_enhanced_scores,
        "Enhanced (Assortative)": final_assortative_scores,
    })

results = simulate_multigenerational_selection_fixed()
print(results.describe().loc[["mean", "50%"]])

Some observers regard genetic selection as an unusual form of human intervention in reproduction. Modern reproductive medicine, however, already involves a wide range of interventions that alter outcomes which would otherwise be determined by chance. In vitro fertilization separates conception from sexual reproduction. Embryo screening permits the identification of chromosomal abnormalities and disease-causing mutations. Prenatal surgery corrects developmental conditions before birth. Neonatal medicine allows the survival of infants who would not have survived in earlier eras.

Concerns that medicine exceeds appropriate human authority have accompanied many medical innovations. Vaccination, anesthesia, organ transplantation, in vitro fertilization, and intensive-care medicine each generated objections that physicians were interfering with natural or divinely ordained processes. Several of these technologies are now routine features of modern healthcare. The recurrence of such objections suggests that debates surrounding reproductive genetics belong to a longer historical pattern accompanying major changes in medical capability.

The use of genetic information in reproductive decision-making predates polygenic scoring. Sperm recipients often request Ivy-league donors; the Repository for Germinal Choice, a genius-level sperm bank that operated from 1980 to 1999 saw large demand [4]. Discussions of embryo selection are also frequently conducted in the context of the history of eugenics. Historically, eugenic movements encompassed a wide range of practices, including immigration restrictions, marriage regulations, compulsory sterilization programs, and voluntary efforts to influence reproduction. Contemporary embryo selection differs in its institutional structure. Rather than directing the reproductive behavior of an entire population, it concerns decisions made by individual families regarding embryos produced through assisted reproduction. As the historian Nathaniel Comfort observes, "The shift from communalism to individual choice has indeed moderated some of the worst abuses" associated with earlier forms of genetic planning.

[3]: though this in particular may take a long time see gwern: https://gwern.net/embryo-selection#overview-of-major-approaches [4]: https://en.wikipedia.org/wiki/Repository_for_Germinal_Choice [5]: https://archive.ph/haBxF#selection-2048.0-2048.1