astir - Automated cell identity from single-cell multiplexed imaging and proteomics

astir is a modelling framework for the assignment of cell type across a range of single-cell technologies such as Imaging Mass Cytometry (IMC). astir is built using pytorch and uses recognition networks for fast minibatch stochastic variational inference.

Key applications:

• Automated assignment of cell type and state from highly multiplexed imaging and proteomic data
• Diagnostic measures to check quality of resulting type and state inferences
• Ability to map new data to cell types and states trained on existing data using recognition neural networks
• A range of plotting and data loading utilities

Getting started

Launch the interactive tutorial:

See the full documentation and check out the tutorials.

Authors

Jinyu Hou, Sunyun Lee, Michael Geuenich, Kieran Campbell

Lunenfeld-Tanenbaum Research Institute & University of Toronto