Inferring modularity in genotype-phenotype maps using patterns of epistasis and pleiotropy

SLED and GTD infer modular genotype–phenotype structure from yeast F1 segregant growth-rate data across 101 environments, yielding highly similar module architectures.

• Why it matters: Inferring low-dimensional genotype–phenotype structure is challenging because loci may act through both epistasis and pleiotropy, yet many approaches treat these signals separately.
• What they did: The authors developed SLED, which models sparse additive and epistatic module contributions, and GTD, which decomposes epistasis per environment into shared basis graphs; they validated both on synthetic data and applied them to yeast growth rates across 101 environments.
• The result: Both models generalize to held-out environments, and compared with SSD, SLED and GTD produce highly similar modular architectures, suggesting consistent genetic architecture across additive and epistatic signals.

AI summary of the abstract · may contain errors · not medical advice · read the paper.