Tutorials

The vignettes are organized around the main choices in a mc-ASTRA analysis: which model to fit, which tissue descriptors to represent, and how to evaluate or interpret the resulting patient map.

Tutorial datasets are available on Zenodo: https://zenodo.org/records/22280030.

For the spatiotemporal dataset, we recommend using the code in the companion repository to generate the data: https://github.com/saezlab/mc-ASTRA_pub.

Using Different Models

Core MOFA workflow

Build an unsupervised multicellular factor model from single-cell data.

Guided sample-level factors with SOFA

Use patient covariates to guide factor discovery and interpretation.

Models across experimental groups

Align datasets and compare multicellular programs across groups.

Pathway-guided factors with MuVI

Use biological prior knowledge to guide feature-level factors.

Spatiotemporal models with MEFISTO

Model temporal structure together with spatial tissue descriptors.

Using Different Tissue Descriptors

Functional views

Represent samples through pathway activities and other functional summaries.

Spatial descriptors

Add neighborhood enrichment and cell-composition views to patient maps.

Multimodal spatial proteomics in CRC

Combine marker intensity, morphology, and spatial feature types.

Evaluation and Downstream Analysis

Integration with patpy and model evaluation

Evaluate patient maps and connect mc-ASTRA outputs to patpy workflows.

Patient archetypes

Identify and interpret extreme tissue states in multicellular factor space.