ParTIpy: a scalable framework for archetypal analysis and Pareto task inference

Trade-offs between different tasks are pervasive across scales in biological systems. For example, cells cannot perform all possible functions simultaneously; instead they allocate limited resources to specialize in subsets of tasks by activating specific gene expression programs. Pareto Task Inference (ParTI) is a framework for analyzing biological trade-offs grounded in multi-objective optimality. However, existing software for ParTI neither scales to large datasets nor integrates well with standard data analysis workflows. To address this gap, we developed ParTIpy ( https://pypi.org/project/partipy ), an open-source Python package that leverages optimization and coreset methods to scale archetypal analysis, the core algorithm underlying ParTI, to millions of cells. By providing tools to characterize archetypes and comprehensive documentation ( https://partipy.readthedocs.io ), ParTIpy integrates seamlessly into existing analysis workflows, especially for single-cell data. We demonstrate how ParTIpy can be used to study intra-cell-type gene expression variability through the lens of task allocation, offering a principled alternative to methods that impose discrete cell state classifications on inherently continuous variation.