Archetypal Analysis network (AAnet)

Quick start

• Guided tutorial for single-cell analysis
• Guided tutorial for simulated simplex
• Guided tutorial for MNIST digits

Introduction

A. Venkat, S. E. Youlten, B. P. San Juan, et al. AAnet resolves a continuum of spatially-localized cell states to unveil tumor complexity. 2024. bioRxiv

Identifying functionally important cell states and structure within a heterogeneous tumor remains a significant biological and computational challenge. Moreover, current clustering or trajectory-based computational models are ill-equipped to address the notion that cancer cells reside along a phenotypic continuum. To address this, we present Archetypal Analysis network (AAnet), a neural network that learns key archetypal cell states within a phenotypic continuum of cell states in single-cell data.