Genos: A Human-Centric Genomic Foundation Model.

Abstract

The rapid expansion of human genomic data demands foundation models that manage ultra-long sequences and capture population diversity, limitations common in existing models which lack human-specific representation and clinical inference efficiency. Here, we introduce Genos (Genos-1.2B/Genos-10B), a human-centric genomic foundation model engineered for million-base-pair sequence modeling. Genos utilizes a large-scale Mixture of Experts (MoE) structure, optimized for a 1Mb context, trained on high-quality human de novo assemblies from datasets such as HPRC and HGSVC, representing diverse global populations. A suite of optimization strategies was implemented to ensure training stability and enhance computational efficiency, which collectively reduces costs and facilitates million-base-pair context modeling. Functionally, Genos performs single-nucleotide resolution analysis and dynamically simulates the cascade effects of non-coding variations on RNA expression profiles. In comprehensive evaluations, Genos uniformly surpasses State-of-the-Art models on critical human genomics benchmarks and demonstrates robust Omics-Text cross-modal diagnostic capabilities. We present a systematic technical evaluation and validation of Genos's architecture, training convergence, and performance across standard benchmarks. This work provides a reliable technical blueprint and performance benchmark for the development of the next generation of high-efficiency genomic foundation models. Genos model weights, inference code, and usage documentation are publicly available on GitHub (https://github.com/BGI-HangzhouAI/Genos) and Hugging Face Hub (https://huggingface.co/BGI-HangzhouAI), with additional cloud services accessible via BGI DCS Cloud-all released under the MIT License.