Transcriptomics
Single-cell and bulk RNA-seq analysis agents
Single-cell and bulk RNA-seq: QC, clustering, differential expression, and pathway analysis
Agents
transcriptomics_expert
The main orchestrator for transcriptomics workflows, coordinating between specialized sub-agents.
Capabilities:
- Workflow planning for scRNA-seq and bulk RNA-seq
- Delegation to specialized sub-agents
- Result synthesis and interpretation
annotation_expert
Specialized agent for single-cell RNA-seq quality control.
Capabilities:
- Cell quality filtering (min genes, min counts)
- Doublet detection (Scrublet integration)
- Mitochondrial content analysis
- Quality metrics visualization
Example workflow:
"Run quality control on my scRNA-seq data with default parameters"de_analysis_expert
Specialized agent for bulk RNA-seq differential expression analysis.
Capabilities:
- Differential expression (DESeq2, pyDESeq2)
- Contrast definition and execution
- Volcano plots and MA plots
- Pathway enrichment integration
Example Workflows
Single-Cell Analysis
User: Load the data from GSE164378 and run a complete single-cell analysis
[transcriptomics_expert delegates to annotation_expert]
- Quality control filtering
- Normalization
- Variable gene selection
- Dimensionality reduction (PCA)
- Clustering (Leiden)
- UMAP visualizationBulk RNA-seq
User: I have bulk RNA-seq data from treated vs control samples.
Run differential expression analysis.
[transcriptomics_expert delegates to de_analysis_expert]
- Sample group assignment
- DESeq2 differential expression
- Volcano plot generation
- Top differentially expressed genes tableDependencies
lobster-transcriptomics requires domain-specific libraries:
| Library | Purpose |
|---|---|
| scanpy | Single-cell analysis framework |
| pydeseq2 | DESeq2 Python implementation |
| scrublet | Doublet detection |
| leidenalg | Community detection for clustering |
These are installed automatically with the package.
Services
lobster-transcriptomics includes domain-specific services bundled with the package:
| Service | Purpose |
|---|---|
| EnhancedSingleCellService | Complete scRNA-seq analysis pipeline (QC, normalization, clustering) |
| QualityService | Cell and gene quality metrics |
| PreprocessingService | Data filtering and normalization |
| DifferentialFormulaService | Formula-based differential expression (DESeq2-style) |
| PseudobulkService | Aggregate single-cell data to pseudobulk |
| BulkVisualizationService | Volcano plots, MA plots, heatmaps for bulk RNA-seq |
| ManualAnnotationService | Manual cell type annotation workflow |
| AnnotationTemplateService | Pre-defined marker gene templates |
Services are installed automatically with the agent package and accessed internally by the agents.
Configuration
Enable transcriptomics agents in your workspace config:
# .lobster_workspace/config.toml
enabled = ["transcriptomics_expert", "annotation_expert", "de_analysis_expert"]Or use a preset:
preset = "scrna-full"Sub-Agent Architecture
transcriptomics_expert (supervisor)
|-- annotation_expert (sub-agent, not directly accessible)
|-- de_analysis_expert (sub-agent, not directly accessible)Sub-agents are accessed through transcriptomics_expert delegation, not directly from the supervisor. This architecture allows the orchestrator to:
- Plan workflows - Determine which sub-agent is needed
- Manage state - Track analysis progress across steps
- Synthesize results - Combine outputs from multiple sub-agents