Information Bottleneck Dominates Adversarial Training for Ancestry-Invariant Polygenic Risk Prediction: Dimensionality, Not Gradient Reversal, Controls the Fairness-Accuracy Tradeoff

In adversarial representation learning for fair prediction, the gradient reversal coefficient ( λ ) is widely treated as the primary control for sensitive-attribute invariance. We show this assumption is wrong. Using a dual-stream architecture for cross-ancestry polygenic risk score (PRS) prediction, we demonstrate that latent dimensionality — the information bottleneck — accounts for 8–27 × more variance in ancestry leakage than adversarial strength. Varying λ across a 20 × range changes leakage by only 2.2 percentage points; varying dimensionality across a 16 × range changes it by 46.6 pp. At dimension 8 with no adversarial training ( λ = 0), ancestry leakage is 32.9% (chance = 20%): the bottleneck alone achieves near-invariance. The adversary architecture (linear vs deep MLP) is equally irrelevant (0.6 pp range). We validate this finding across two unrelated domains — genomic ancestry invariance (6 clinical traits, 1000 Genomes, n = 2,504) and EEG subject invariance (pretrained HFTP + Braindecode dual-domain model, 20 subjects) — observing consistent dimensionality dominance (12.7:1 ratio in EEG).

For the genomic application, Stream 1 encodes population structure via DCT-II frequencydomain features (136 coefficients); Stream 2 encodes phenotype signal from top PRS SNPs (PCA to 128 dimensions). The architecture works equally well with standard genomic PCA as the ancestry stream ( R ² = 0.217 vs 0.222), confirming the contribution is architectural, not encoding-specific. African-ancestry PRS reconstruction R ² improves on all six traits (e.g., +5.1 pp for coronary artery disease). Linear models achieve higher aggregate R ² but fail catastrophically on cross-ancestry transfer ( R ² = − 12.45 for African-ancestry CAD). We emphasize that we predict PRS (a computed score), not disease phenotypes; validation on biobank-scale phenotype data is ongoing.

These results suggest the adversarial fairness community has been over-investing in adversary engineering relative to simple capacity control. Practitioners should select latent dimensionality first to set the information budget for the fairness-accuracy tradeoff, then optionally use adversarial training for marginal refinement.