Agent-Skills.md

Agent Skills: Developing-Heart-Local Skill

Developing Heart Atlas notebooks converted to HTML (local)

UncategorizedID: Ketomihine/my_skills/developing-heart-local

Repository

Ketomihine/my_skills
Ketomihine

Install this agent skill to your local

pnpm dlx add-skill https://github.com/Ketomihine/my_skills/tree/HEAD/developing-heart-local

Skill Files

Browse the full folder contents for developing-heart-local.

Download Skill

Loading file tree…

developing-heart-local/SKILL.md

Skill Metadata

Name
developing-heart-local
Description
Developing Heart Atlas notebooks converted to HTML (local)

Developing-Heart-Local Skill

Comprehensive assistance with developing-heart-local development, generated from official documentation.

When to Use This Skill

This skill should be triggered when:

  • Working with fetal heart development single-cell analysis
  • Processing Xenium spatial transcriptomics data for cardiac tissue
  • Performing cell-type annotation and clustering on cardiac cell populations
  • Analyzing ventricular cardiomyocytes and conduction system cells
  • Working with Scanpy for heart development transcriptomics
  • Implementing harmony batch correction for cardiac datasets
  • Creating UMAP embeddings and visualizations for cardiac cell data
  • Performing differential expression analysis in heart development
  • Working with leiden clustering for cardiac cell populations

Quick Reference

Essential Setup and Data Loading

Example 1: Import Required Libraries

import matplotlib.pyplot as plt
import scanpy as sc
import numpy as np
import pandas as pd
import seaborn as sns
import os
import matplotlib as mpl

# Configure matplotlib for proper text rendering
mpl.rcParams['pdf.fonttype'] = 42
mpl.rcParams['ps.fonttype'] = 42
import scanpy.external as sce

Example 2: Configure Scanpy Settings

sc.settings.set_figure_params(dpi=80)

Example 3: Load Cardiac Single-Cell Data

adata_dir = '/home/kk837/rds/rds-teichlab-C9woKbOCf2Y/kk837/Foetal/anndata_objects/Xenium/subsets'
sample_id = 'C194-HEA-0-FFPE-1_Hst45-HEA-0-FFPE-1_concat'
celltype = 'VentricularCardiomyocytes'

# Read in the data
adata = sc.read_h5ad(f'{adata_dir}/{sample_id}_5K_{celltype}_lognorm.h5ad')

Data Visualization and Exploration

Example 4: Create UMAP Visualization

latent_space = 'pca_harmony'
sc.pl.embedding(adata, basis=f"X_umap_{latent_space}",
                color=['total_counts','n_genes_by_counts','cell_area','tissue_block_id'],
                wspace=0.2, cmap='RdPu', vmax='p100')

Example 5: Visualize Cell-Type Annotations

sc.pl.embedding(adata, basis=f"X_umap_{latent_space}",
                color=['multi_celltypes_coarse',"conf_score_midmod2fine","celltypist_midmod2fine"],
                wspace=0.2, cmap='RdPu', vmax='p100')

Clustering Analysis

Example 6: Perform Leiden Clustering

resolutions_list = [0.2, 0.4, 0.6, 1, 1.2, 1.5, 2]
for resolution in resolutions_list:
    sc.tl.leiden(adata, neighbors_key=latent_space,
                 resolution=resolution, key_added=f'leiden_{str(resolution)}', n_iterations=2)

Example 7: Visualize Clustering Results

sc.pl.embedding(adata, basis=f"X_umap_{latent_space}",
                color=[f'leiden_{str(resolution)}' for resolution in resolutions_list],
                wspace=0.3)

Marker Gene Analysis

Example 8: Define Cardiac Cell-Type Markers

markers_others = {
    'CM': ['TTN', 'TNNT2','MYH6', 'MYH7', 'MYL2', 'FHL2', 'NPPA', 'MYL7', 'MYL4'],
    'FB': ['DCN', 'GCN', 'PDGFRA','COL1A1','COL1A2'],
    'EC': ['VWF', 'PECAM1', 'CDH5','RGCC', 'FABP5'],
    'Peri': ['RGS5', 'ABCC9', 'KCNJ8'],
    'SMC': ['MYH11', 'TAGLN', 'ACTA2'],
    'Neuro': ['PLP1', 'NRXN1', 'NRXN3','PRPH', 'NEFL', 'NEFM', 'NEFH', 'STMN2'],
    'Myelo': ['CD14', 'C1QA', 'CD68','LYVE1','TIMD4'],
}

# Filter markers to only include genes present in the dataset
for key in markers_others.keys():
    markers_others[key] = [x for x in markers_others[key] if x in adata.var_names]

Example 9: Create Dotplot for Marker Genes

resolution_sel = 1
sc.tl.dendrogram(adata, groupby=f'leiden_{str(resolution_sel)}')
sc.pl.dotplot(adata,
              markers_others,
              groupby=f'leiden_{str(resolution_sel)}',
              dendrogram=True,
              standard_scale="var",
              color_map="Reds",
              swap_axes=False)

Cell-Type Annotation

Example 10: Manual Cell-Type Assignment

# Create manual annotations based on clustering
adata.obs['tmp_annotation'] = adata.obs[f'leiden_{str(resolution_sel)}'].copy()
adata.obs.replace({'tmp_annotation':{
    '0':'VentricularCardiomyocytes',
    '1':'unassigned',
    '2':'VentricularCardiomyocytes',
    '3':'VentricularCardiomyocytes',
    '4':'VentricularCardiomyocytes',
    '5':'VentricularCardiomyocytesCycling',
    '6':'VentricularCardiomyocytes',
    '7':'VentricularCardiomyocytes',
    '8':'VentricularCardiomyocytes',
    '9':'VentricularCardiomyocytes',
    '10':'VentricularCardiomyocytes',
}}, inplace=True)

Reference Files

This skill includes comprehensive documentation in references/:

  • notebooks.md - Complete collection of Jupyter notebooks converted to HTML, containing:
    • Step-by-step cardiac data processing workflows
    • Cell-type annotation protocols for ventricular cardiomyocytes
    • Integration and clustering methodologies
    • Marker gene identification and validation
    • Spatial transcriptomics analysis pipelines
    • Session information and package versions for reproducibility

The notebooks documentation includes:

  • 209 pages of detailed cardiac single-cell analysis workflows
  • Real code examples with proper Python syntax highlighting
  • Data visualization techniques for cardiac development studies
  • Practical guidance on Xenium data processing
  • Manual annotation strategies for cardiac cell subtypes

Working with This Skill

For Beginners

Start with the basic data loading and visualization examples:

  1. Set up your environment with the required libraries (Example 1)
  2. Learn to load cardiac single-cell data (Example 3)
  3. Master basic UMAP visualizations (Example 4)
  4. Understand the structure of cardiac AnnData objects

For Intermediate Users

Focus on clustering and marker analysis:

  1. Implement leiden clustering at multiple resolutions (Example 6)
  2. Create comprehensive marker gene plots (Examples 8-9)
  3. Learn manual cell-type annotation strategies (Example 10)
  4. Explore harmony batch correction for cardiac datasets

For Advanced Users

Dive into specialized cardiac analysis:

  1. Work with ventricular cardiomyocyte subpopulations
  2. Analyze conduction system cell markers
  3. Implement spatial transcriptomics analysis
  4. Perform differential expression in cardiac development contexts

Navigation Tips

  • Use the latent_space variable consistently throughout your analysis
  • Always check gene presence in adata.var_names before analysis
  • Leverage the extensive marker libraries provided in the examples
  • Use session_info.show() for reproducibility documentation

Key Concepts

Core Terminology

  • VentricularCardiomyocytes: Main contractile cells of the heart ventricles
  • Xenium: Spatial transcriptomics platform for high-resolution mapping
  • Harmony: Batch correction method for integrating multiple cardiac datasets
  • Leiden clustering: Community detection algorithm for cell population identification
  • UMAP: Dimensionality reduction technique for visualizing high-dimensional cardiac data

Data Structure

The cardiac AnnData objects typically contain:

  • obs: Cell metadata including counts, area, and cell-type annotations
  • var: Gene information with highly variable gene analysis
  • obsm: Dimensionality reductions (PCA, Harmony, UMAP) and spatial coordinates
  • uns: Analysis results including color palettes and clustering parameters

Analysis Pipeline

  1. Data Loading: Import Xenium cardiac data with proper preprocessing
  2. Quality Control: Assess transcript counts and cell area metrics
  3. Dimensionality Reduction: PCA with Harmony batch correction
  4. Clustering: Multi-resolution leiden clustering for population discovery
  5. Marker Identification: Use cardiac-specific gene panels for annotation
  6. Visualization: UMAP plots with cardiac-relevant color schemes

Resources

references/

The notebooks documentation provides:

  • Complete cardiac analysis workflows from raw data to final annotations
  • Reproducible code examples with package version information
  • Detailed explanations of cardiac cell-type markers and their biological significance
  • Step-by-step guidance for spatial transcriptomics analysis

scripts/

Add helper scripts for:

  • Cardiac marker gene validation
  • Batch correction optimization
  • Spatial visualization enhancement
  • Cell-type annotation automation

assets/

Include:

  • Cardiac marker gene libraries
  • Color palettes optimized for cardiac data visualization
  • Template notebooks for new cardiac projects
  • Reference datasets for benchmarking

Notes

  • This skill specializes in fetal heart development and cardiac single-cell analysis
  • All examples are derived from real cardiac research workflows
  • The documentation emphasizes ventricular cardiomyocyte analysis and conduction system studies
  • Spatial transcriptomics integration is a key feature of this skill
  • Code examples prioritize reproducibility with session information tracking

Updating

To refresh this skill with updated cardiac analysis methodologies:

  1. Re-run the documentation scraper with the same cardiac-focused configuration
  2. The skill will be rebuilt with the latest cardiac research workflows and best practices
  3. New cardiac marker discoveries and analysis techniques will be automatically integrated