omics-analysis-skills-index

id	omics_skills_index
name	Omics Analysis Skills Index
description	Skills for single-cell and spatial omics data analysis. Best practices, code snippets, and workflows for the scverse ecosystem.

Agent Skills for Omics Data Analysis

Best practices and workflows for single-cell and spatial omics analysis. Load the relevant skill files when performing specific analysis tasks.

Core Single-Cell Skills

High-priority, actionable workflows for the most common single-cell analysis tasks.

Skill index: single_cell/SKILL.md

Skills:

• Quality Control: Filtering, doublet detection, normalization, QC metrics
• Cell Type Annotation: Marker-based and reference-based label assignment
• Trajectory Inference: Pseudotime, lineage tracing, RNA velocity

---

Gene Panel Selection

End-to-end workflow for designing gene panels in scRNA-seq and spatial transcriptomics (HVG/DE/RF/scGeneFit/SpaPROS), with sub-panel discovery, consensus scoring, biological completion, and benchmarking.

Skill folder: gene_panel_selection/

When to use:

• Designing a gene panel for spatial transcriptomics in a static / terminal system
• Benchmarking existing panels (ARI/NMI/Silhouette + UMAP)
• IMPORTANT: When doing gene panel selection, strictly follow this workflow

---

Developmental Gene Panel Design

Panel design for developing / dynamic systems (embryonic organs, differentiation, regeneration), where terminal cell types are end-products of earlier lineage programs. Uses TWO references — the target (usually late) stage for cell-state resolution, and an independent EARLIER-stage reference mined by trajectory/pseudotime ranking for developmental regulators — under an explicit budget split.

Skill folder: developmental_gene_panel/

When to use (use this INSTEAD of gene_panel_selection):

• The tissue is embryonic / differentiating / regenerating, or the user cares about lineage origin and regulators, not only terminal cell-type classification
• The assay is at a late stage but the biology is a trajectory
• Why it matters: developmental regulators are frequently NOT recoverable from the target stage — in a benchmarked case study six unbiased methods (cell-type DE, DPT pseudotime Pearson/Spearman, cross-stage ANOVA, global marker rank) all ranked a known CHD regulator in the top 0.6–1.8% on an early reference and failed (top 3–79%) on the target stage; literature lookup also failed

---

Spatial Omics

Skills for spatial transcriptomics mapping, imputation, and 3D visualization.

Skill index: spatial/SKILL.md

Skills:

• Single-Cell to Spatial Mapping: Map scRNA-seq to spatial data with MOSCOT for gene imputation and cell type transfer
• 3D Spatial Visualization: Interactive 3D plots and rotating animations with PyVista

When to use:

• You have paired scRNA-seq and spatial transcriptomics data
• You want to impute genes or transfer cell type labels to spatial coordinates
• Your spatial data has 3D coordinates and you want to visualize them

---

Single-Cell Foundation Models (SCFM)

Workflow and model reference for embedding/integration with foundation models (scGPT, Geneformer, UCE, scBERT, etc.).

Skill index: scfm/SKILL.md

When to use:

• You want FM embeddings (e.g., obsm["X_uce"], obsm["X_scGPT"])
• You need model selection based on gene ID scheme and species
• You want a validation-first workflow before heavy inference

---

Database Access

Tools for querying genomic databases, downloading sequencing data, and accessing large-scale single-cell datasets programmatically.

Skill index: database_access/SKILL.md

Tools covered:

• gget: 23 modules for querying Ensembl, NCBI, UniProt, COSMIC, OpenTargets, etc.
• iSeq: CLI for downloading from GSA, SRA, ENA, DDBJ, GEO
• CZ CELLxGENE Census: API for 217M+ single-cell observations

---

Upstream Processing

Technology-specific pipelines for processing raw sequencing data into analysis-ready count matrices.

Skill index: upstream_processing/SKILL.md

Technologies covered:

• nf-core Pipelines: 143+ Nextflow pipelines for scRNA-seq, spatial, bulk, ATAC-seq, ChIP-seq, variant calling
• OpenST: Open-source spatial transcriptomics processing pipeline

---

General Data Analysis

Cross-cutting skills for environment setup and computational performance.

Skill index: general_data_analysis/SKILL.md

Skills:

• Environment Management: Conda/Mamba/venv setup for reproducible environments
• Parallel Computing: Multi-core CPU, GPU acceleration, memory optimization

---

Supplementary Reference: SC Best Practices

Comprehensive guidance derived from the Single-cell Best Practices book. Use as supplementary context when the core skills above need deeper background.

Skill index: sc_best_practices/SKILL.md

Topics covered:

• Preprocessing, normalization, dimensionality reduction
• Clustering, annotation, dataset integration
• Trajectory analysis, RNA velocity, lineage tracing
• Differential expression, compositional analysis, pathway analysis
• Gene regulatory networks, cell-cell communication
• Bulk deconvolution, scATAC-seq, spatial omics
• CITE-seq, immune repertoire (TCR/BCR)
• Multimodal integration, reproducibility

---

Using Skills

1. Before analysis: Scan this index for relevant skills
2. Load skill file: Read the full skill document for detailed guidance
3. Follow best practices: Use the code snippets and workflows provided
4. Adapt as needed: Skills are templates; adjust for your specific data