Session 05 · No-Code & Agentic AI for Life Sciences

Omics Data
Analysis: Single-Cell RNA-Seq

Cell-level resolution · Clustering · Trajectories · Communication

Md. Jubayer Hossain · Founder & CEO
DeepBio Limited · DeepBio Academy

No Code & Agentic AI for Life Sciences · Session 05 · June 2026

Pipeline Overview

scRNA-seq No-Code Workflow

FASTQ
Cell Ranger
QC
Filtering
DIM RED
PCA/UMAP
CLUSTER
Cell Types

🧠 Seurat / Scanpy

The two dominant frameworks for single-cell analysis. While typically code-based, we use agent-assisted execution to run these pipelines without writing syntax.

📦 The Deliverable

An AnnData (h5ad) or Seurat Object containing expressions, embeddings, and metadata for thousands of individual cells.

Pre-processing

Quality Control & Cell Filtering

Removing "noise" is critical for single-cell success. We filter based on three primary metrics:

  • nFeature_RNA: Number of unique genes (detects empty drops or doublets).
  • nCount_RNA: Total molecules (depth).
  • percent.mt: Mitochondrial percentage (high % indicates dying cells).

🎯 Thresholding

Typically: < 5% mt-DNA, > 200 genes/cell. Agents can suggest optimal cutoffs by analyzing violin plots.

"QC is not just about cleaning data; it's about preserving biological signal while removing technical artifacts."
Manifold Learning

Integration & Clustering

🤝 Batch Integration

Correcting technical differences between samples (Harmony, LIGER) to ensure cells cluster by biology, not by date of experiment.

🧩 Graph-Based Clustering

Louvain or Leiden algorithms identify groups of similar cells. Visualized via UMAP or t-SNE for 2D interpretation.

Question: How do we know what these clusters actually represent? ➜ Cell Type Identification.

Annotation

Cell Type Identification

Assigning biological labels to clusters using marker genes.

  • FindAllMarkers: Identifies top genes per cluster.
  • Reference Mapping: Using SingleR or Azimuth to map your cells to a known atlas.
  • Manual Curation: Comparing marker lists to canonical literature.

🤖 Agentic Annotation

"Review this list of markers: CD3D, CD4, IL7R. What cell type is this?"

Agent: "These are canonical markers for CD4+ T Cells."

Downstream Analysis

Proportions & Pathway Analysis

📊 Proportion Testing

Is the percentage of Macrophages significantly higher in the 'Treated' group vs 'Control'?

🧬 Cell-Specific DEG

Differential expression within a single cell type across conditions (e.g., Exhausted vs. Naive T cells).

🌈 GSEA / Pathway

Functional enrichment for specific clusters to understand their unique metabolic or signaling state.

Dynamic States

Trajectory & Pseudotime Analysis

Cells are dynamic. Trajectory analysis (Monocle3, Slingshot) models cells as a continuous process.

  • Pseudotime: A measure of how far a cell has progressed along a differentiation path.
  • Branching: Identifying points where cells "choose" different fates.
State A ➜ State B ➜ State C

Modeling cellular transitions and fate decisions.

Interactomics

Cell-Cell Communication

Using **CellChat** or **CellPhoneDB** to predict how cell types talk to each other via Ligand-Receptor pairs.

  • Sender vs. Receiver: Who is producing the signal? Who has the receptor?
  • Strength: Quantifying the probability of interaction.
  • Pathways: Identifying active signaling axes (e.g., TGF-beta, WNT).

💡 Insight

Crucial for understanding the tumor microenvironment or immune response, where interaction is as important as expression.

Predictive Biology

RNA Velocity Analysis

Predicting the **future state** of individual cells by comparing Spliced vs. Unspliced mRNA (scVelo).

  • Positive Velocity: Gene is being actively upregulated.
  • Negative Velocity: Gene is being downregulated.
  • Streamlines: Visualizing the "flow" of cells across the UMAP.

🚀 Real-Time Direction

Unlike pseudotime (which is a relative ordering), RNA velocity provides the actual biological vector of change.

Summary

Session 05 Key Takeaways

  • scRNA-seq provides cell-level granularity.
  • QC (mtDNA, doublets) is the foundation of trust.
  • Integration ensures biology overrides batch effects.
  • Annotation combines auto-mapping with AI marker review.
  • Trajectories model cellular "time" and fate.
  • Communication analysis reveals signaling networks.
  • RNA Velocity predicts the future state of the tissue.
  • AI agents automate the heavy lifting of scRNA-seq.

Next Session: Prepare Claude for AI Workflows — Building your custom research assistant.

No Code & Agentic AI for Life Sciences

Thank You

Get in Touch

Md. Jubayer Hossain

bio.link/hossainlab

© 2026 Md. Jubayer Hossain · No Code & Agentic AI for Life Sciences — Session 05