pyaging: a Python-based compendium of GPU-optimized aging clocks.

Area of Science:

• Biogerontology
• Computational Biology
• Bioinformatics

Background:

• Aging is linked to diseases and mortality, with molecular changes offering potential biomarker development.
• Machine learning models, known as aging clocks, are used to predict biological age.
• A lack of robust, Python-based software hinders the integration and comparison of diverse aging clock models.

Purpose of the Study:

• To introduce pyaging, an open-source Python package for comprehensive aging research.
• To address the need for integrated and comparable aging clock models.
• To provide a versatile tool for analyzing molecular data across species.

Main Methods:

• Developed pyaging, a Python package harmonizing dozens of aging clocks.
• Integrated support for diverse molecular data types (DNA methylation, transcriptomics, ChIP-Seq, ATAC-Seq).
• Implemented a PyTorch-based backend for GPU acceleration and rapid inference.
• Enabled multi-species analysis (human, mammals, C. elegans).

Main Results:

• pyaging harmonizes numerous aging clocks across various molecular data types.
• The package supports a wide range of model types, including linear, PCA, neural networks, and ARO models.
• GPU acceleration ensures rapid inference for large datasets and complex models.
• Multi-species analysis capability enhances its broad applicability in aging research.

Conclusions:

• pyaging provides a robust, open-source solution for integrating and comparing diverse aging clock models.
• The package facilitates advanced aging research by offering efficient data analysis and multi-species support.
• pyaging is readily available on GitHub, PyPI, and Zenodo for the research community.