User Guide

---

Setting up

To use MIRA, you need a single-cell multiomics, scRNA-seq, or scATAC-seq experiment preprocessed and formatted as a peaks by cells count matrix for scATAC-seq and genes by cells count matrix for scRNA-seq. Before starting with MIRA, make sure to filter low quality cells, doublets, etc.

These tutorials also depend on scanpy. To install:

$ conda install -c conda-forge -c bioconda scanpy

Tutorials

The following tutorials cover the main functionalities of MIRA, starting from initial steps - topic modeling - and culminating in the identification of divergences between local chromatin accessibility and gene expression across genes and cell states.

• Topic Modeling with MIRA+CODAL
  • What is a topic model?
  • Tuning and training
  • A note on data modalities
  • Step 1: Feature selection and Preprocessing - Gene Expression
  • Step 2: Model instantiation
  • Step 3: Hyperparameter Optimization
  • Step 4: Analysis
  • Assessing disentanglement
  • Model persistence
• Atlas-level integration
  • Preprocessing ATAC-seq data
  • Caching data to disk
  • Choosing num topics
  • (Optional) Refining the model using Bayesian optimization
• Joint Representation
  • Predicting topics
  • Joining modalities
  • Analyzing joint topic compositions
  • The box-cox parameter
• Topic Analysis
  • Expression Topic Analysis
  • Accessibility Topic Analysis
  • Motif scoring
  • Saving analyses
• Pseudotime Trajectory Inference
  • Defining pseudotime
  • Lineage inference
  • Visualizing lineages and topics
  • Transport map tracing
• Regulatory potential modeling
  • TSS annotations
  • RP model training
  • Defining local chromatin neighborhoods
  • Predicting expression from accessibility
  • Gene-TF targeting
  • Visualizing RP models
• NITE regulation
  • Training NITE models
  • Chromatin differential
  • NITE scores
• Conclusion and future directions

Other Features

These tutorials cover functionalities of MIRA that are useful, but outside of the main analysis track. Learn how to make beautiful and informative streamgraphs and use MIRA plotting functions with velocity data.

• Plotting Streams
  • Stream mode: Plotting Topics
  • Line mode: Comparing modes
  • Swarm mode: Discrete features
  • Split mode: Visualizing marker genes
  • Heatmap mode: many features
  • “Zooming” in on streams
  • Scatter mode: traditional plots
  • Next
• Using Velocity and Cellrank with Streams
  • ScVelo/CellRank workflow
  • Lineage parsing with MIRA
• Enrichr Tutorial
  • Example
  • Functions
• Assembling TSS annotations

Examples

We will add more examples of MIRA analyses in the future! These examples may be run in Google Colab for an interactive experience.

• 10X Mouse Brain Example

previous

Getting Started

next

Topic Modeling with MIRA+CODAL