Overview

PyRHE is a unified and efficient Python framework for genomics heritability estimation. It provides a modular and extensible platform for implementing various genetic architecture estimation models, while providing computation optimizations for large-scale genomic data.

Computational Efficiency

PyRHE integrates two core optimizations for computational efficiency:

Parallel Processing

• Jackknife subsamples are tiled and distributed across multiple workers

• Parallelizes genotype I/O and matrix computations

• Efficient memory management through shared memory arrays

Tensor Operations

• Large matrices automatically converted to tensors using PyTorch

• Optimized matrix multiplications for both CPU and CUDA-enabled GPU architectures

Unified Framework

Base Classes

The framework is built around two core abstract classes:

• Base: Provides common infrastructure for all models

• StreamingBase: Extends Base for memory-efficient processing

Key features of the base classes:

• Abstract interfaces for model-specific implementations

• Shared infrastructure for data processing and estimation

• Built-in support for jackknife resampling

• Efficient multiprocessing and memory management

Model Support and Extensibility

The framework natively supports multiple models:

• RHE

• RHE-DOM

• GENIE

Each model shares common infrastructure while implementing model-specific components through well-defined interfaces. In addition, the framework is designed for easy extension to include new models.