Reverse Rain (RRain)

RRain is a visualization framework for representing single-cell RNA-seq data in functional space using protein domains.

It extends the PLANT framework by projecting cells into a low-dimensional embedding and stacking their domain-level features vertically, creating “towers” that resemble upside-down rain emerging from a cloud of cells.

Instead of genes defining the primary structure, RRain aggregates expression into protein domains (e.g., Pfam)—allowing visualization of functional composition per cell.

---

🌐 Demo

https://asifali-bio.github.io/rrain/

---

Concept

RRain separates the data into three intuitive dimensions:

• (x, y) → latent embedding of cells (domain space PCA/UMAP)
• z → feature index (protein domains)
• size → domain-level expression (aggregated counts / TPM)
• color → cell clusters (e.g., Seurat clustering)

Each cell becomes a vertical tower of domains, positioned according to similarity in functional space.

---

Key idea

Traditional scRNA-seq pipelines operate in gene space, using highly variable genes for dimensionality reduction and clustering.

RRain instead constructs a parallel representation in domain space:

cells × genes → cells × domains

by mapping:

gene → protein → Pfam domain

This enables:

• Functional aggregation across genes
• Reduced redundancy (shared domains)
• Sharper boundaries between cell states
• Visualization of transitions via domain composition

---

What it shows

RRain allows simultaneous visualization of:

• Global similarity between cells (embedding)
• Functional structure of each cell (tower composition)
• Domain presence/absence (gaps in towers)
• Domain magnitude (bubble size)
• Cluster identity (color via Seurat)

---

Workflow

1. Input data

   • scRNA-seq count matrix (cells × genes)

2. Domain annotation

   • Run InterProScan on a reference proteome
   • Extract Pfam domains
   • Map UniProt → gene symbols (Ensembl / biomaRt)

3. Aggregation

   • Map genes → domains
   • Sum expression per domain per cell

cells × genes → cells × domains

4. Embedding

• PCA / UMAP

5. Visualization

• Convert to long format

• Plot 3D towers:

  • x, y = embedding
  • z = domain index
  • size = domain expression
  • color = cluster identity

---

Relationship to PLANT

RRain builds directly on PLANT:

• PLANT → bulk / cross-sample domain composition
• RRain → single-cell domain composition

RRain extends the idea from:

“one tower per sample”

to:

“one tower per cell”

enabling exploration of cellular heterogeneity and transitions.

---

Example use cases

• Single-cell RNA-seq (primary use case)
• Functional profiling of cell states
• Domain-level clustering and boundary detection
• Comparative analysis of cell populations

---

🔧 Prototype stage

• Current implementation uses point-based towers (Plotly)

---

Banner emoji designed by OpenMoji.