The MERGE (Model Ensemble for Ranking Genomic Effect) framework is a cutting-edge pipeline designed for variant pathogenicity prediction. By avoiding reliance on clinical databases that cause "data circularity," MERGE utilizes a Dynamic Shunting Architecture to categorize variants into coding, non-coding, and splice regions.
The system integrates powerful foundation models, including Evo, HyenaDNA, Nucleotide Transformer, and AlphaGenome, ensuring robust performance across diverse genomic landscapes.
We provide a professional CLI tool for high-throughput analysis. Officially published on PyPI, it supports complete offline execution and GPU-accelerated prediction.