OmeSim: a genetics-based nonlinear simulator for in-between-ome and phenotype

Abstract Deciphering the genetic basis of complex traits increasingly leverages intermediate molecular layers (“in-between-omes,” e.g., the transcriptome) in association studies such as transcriptome-wide association studies. Despite many emerging statistical and machine-learning approaches, there is no flexible standard for simulating phenotypes from genotypes while explicitly modeling the role of an in-between-ome, especially under nonlinear architectures (e.g., co-expression networks). This gap hampers fair power estimation and rigorous benchmarking. We present OmeSim, a configurable simulator that jointly generates genotype, an in-between-ome, and phenotype, capturing complex (including nonlinear) relationships. OmeSim outputs the full generative/causal graph together with data matrices and the induced correlation and association structures, providing gold-standard datasets for developing and evaluating methods that integrate an in-between-ome into genotype–phenotype studies. We validate OmeSim by comparing simulated human transcriptomes to real human transcriptomes and by benchmarking alternative association-mapping tools, demonstrating its utility for reproducible power analyses and method comparison in multi-omics integration. Source code and documentation: https://github.com/zhoulongcoding/OmeSim.