This source did not publish a separate summary. Review SKILL.md before using the skill.
SKILL.md
Multi-Omics Integration
Coordinate and integrate multiple omics datasets for comprehensive systems biology analysis. This skill orchestrates specialized ToolUniverse skills to perform cross-omics correlation, multi-omics clustering, pathway-level integration, and unified interpretation across molecular layers.
When to Use This Skill
Triggers:
User has multiple omics datasets (RNA-seq + proteomics, methylation + expression, etc.)
Requests for integrative multi-omics analysis
Cross-omics correlation queries (e.g., "How does methylation affect expression?")
Multi-omics biomarker discovery
Systems biology questions requiring multiple molecular layers
Precision medicine applications with multi-omics patient data
Questions about molecular mechanisms across omics types
Example Questions This Skill Solves:
"Integrate RNA-seq and proteomics data to find genes with concordant changes"
"How does promoter methylation correlate with gene expression?"
"Perform multi-omics clustering to identify patient subtypes"
"Which pathways are dysregulated across transcriptome, proteome, and metabolome?"
"Find multi-omics biomarkers for disease classification"
"Correlate CNV with gene expression to identify dosage effects"
"Integrate GWAS variants, eQTLs, and expression data"
"Perform MOFA+ analysis on multi-omics cancer data"
Core Capabilities
Capability
Description
Data Integration
Match samples across omics, handle missing data, normalize scales
Cross-Omics Correlation
Correlate features across molecular layers (gene expression vs protein, methylation vs expression)
Multi-Omics Clustering
MOFA+, NMF, joint clustering to identify omics-driven subtypes
Pathway Integration
Combine omics evidence at pathway level for unified biological interpretation
Biomarker Discovery
Identify multi-omics signatures with improved predictive power
Objective: Identify common samples across omics datasets.
Sample ID harmonization:
def match_samples_across_omics(omics_data_dict):
"""
Match samples across multiple omics datasets.
Parameters:
omics_data_dict: {
'rnaseq': DataFrame (genes x samples),
'proteomics': DataFrame (proteins x samples),
'methylation': DataFrame (CpGs x samples),
'cnv': DataFrame (genes x samples)
}
Returns:
- common_samples: List of sample IDs present in all omics
- matched_data: Dict of DataFrames with common samples only
"""
# Extract sample IDs from each omics
sample_ids = {
omics_type: set(df.columns)
for omics_type, df in omics_data_dict.items()
}
# Find common samples (intersection)
common_samples = set.intersection(*sample_ids.values())
# Subset each omics to common samples
matched_data = {
omics_type: df[sorted(common_samples)]
for omics_type, df in omics_data_dict.items()
}
return sorted(common_samples), matched_data
Handling missing omics:
Pairwise integration if not all samples have all omics
Document sample availability matrix
Phase 3: Feature Mapping
Objective: Map features from different omics to common gene-level identifiers.
Gene-centric integration:
# Map all features to genes
feature_mapping = {
'rnaseq': 'gene_symbol', # Already gene-level
'proteomics': 'gene_symbol', # Map protein to gene
'methylation': 'gene_symbol', # Map CpG to gene (promoter)
'cnv': 'gene_symbol', # CNV regions to overlapping genes
'metabolomics': 'enzyme_gene' # Metabolite to enzyme gene
}
CpG to gene mapping:
Promoter methylation: CpGs within TSS ± 2kb
Gene body methylation: CpGs within gene boundaries
Average methylation per gene (weighted by probe coverage)
CNV to gene mapping:
Use variant-analysis skill to identify genes in CNV regions
Calculate copy number per gene (log2 ratio)
Phase 4: Cross-Omics Correlation
Objective: Correlate features across molecular layers to understand regulation.
Example analyses:
4.1: Expression vs Protein (Translation Efficiency)
def correlate_rna_protein(rnaseq_data, proteomics_data):
"""
Correlate mRNA and protein levels for each gene.
Expected: Positive correlation (r ~ 0.4-0.6 typical)
Discordance indicates post-transcriptional regulation
"""
# Find common genes
common_genes = set(rnaseq_data.index) & set(proteomics_data.index)
correlations = {}
for gene in common_genes:
rna = rnaseq_data.loc[gene]
protein = proteomics_data.loc[gene]
# Spearman correlation (robust to outliers)
r, p = spearmanr(rna, protein)
correlations[gene] = {'r': r, 'p': p}
# Identify discordant genes (low RNA-protein correlation)
discordant = {g: v for g, v in correlations.items() if abs(v['r']) < 0.2}
return correlations, discordant
4.2: Methylation vs Expression (Epigenetic Regulation)
def correlate_methylation_expression(methylation_data, rnaseq_data):
"""
Correlate promoter methylation with gene expression.
Expected: Negative correlation (increased methylation → decreased expression)
"""
# For each gene with promoter methylation
results = {}
for gene in methylation_data.index:
if gene in rnaseq_data.index:
meth = methylation_data.loc[gene] # Average promoter beta
expr = rnaseq_data.loc[gene]
r, p = spearmanr(meth, expr)
results[gene] = {'r': r, 'p': p, 'direction': 'repressive' if r < 0 else 'activating'}
# Identify genes with strong methylation-expression anticorrelation
regulated = {g: v for g, v in results.items() if v['r'] < -0.5 and v['p'] < 0.01}
return results, regulated
4.3: CNV vs Expression (Dosage Effect)
def correlate_cnv_expression(cnv_data, rnaseq_data):
"""
Correlate copy number with gene expression.
Expected: Positive correlation (gene dosage effect)
"""
results = {}
for gene in cnv_data.index:
if gene in rnaseq_data.index:
cnv = cnv_data.loc[gene] # log2 ratio
expr = rnaseq_data.loc[gene]
r, p = pearsonr(cnv, expr)
results[gene] = {'r': r, 'p': p}
# Genes with dosage effect (CNV drives expression)
dosage_genes = {g: v for g, v in results.items() if v['r'] > 0.5 and v['p'] < 0.01}
return results, dosage_genes
Phase 5: Multi-Omics Clustering
Objective: Identify patient subtypes using integrated omics data.
Method 1: MOFA+ (Multi-Omics Factor Analysis)
MOFA+ identifies latent factors that explain variation across omics.
# Conceptual workflow (uses R's MOFA2 package or Python implementation)
# 1. Prepare multi-omics data as list of matrices
# 2. Run MOFA+ to identify factors
# 3. Inspect factor variance explained per omics
# 4. Cluster samples based on factor scores
# Example interpretation:
# Factor 1: Explains 40% variance in RNA-seq, 30% in proteomics → Cell proliferation
# Factor 2: Explains 50% variance in methylation → Epigenetic subtype
# Factor 3: Explains 20% variance in CNV → Genomic instability
Decompose multi-omics matrices into shared latent components.
def joint_nmf_clustering(omics_data_dict, n_clusters=3):
"""
Perform joint NMF across omics for clustering.
Returns patient cluster assignments based on shared factors.
"""
# Concatenate omics matrices (after normalization)
combined_matrix = np.vstack([
omics_data_dict['rnaseq'].values,
omics_data_dict['proteomics'].values,
omics_data_dict['methylation'].values
])
# Run NMF
from sklearn.decomposition import NMF
model = NMF(n_components=n_clusters, init='nndsvd', random_state=42)
W = model.fit_transform(combined_matrix) # Feature loadings
H = model.components_ # Sample coefficients
# Cluster samples based on H (components)
from sklearn.cluster import KMeans
clusters = KMeans(n_clusters=n_clusters).fit_predict(H.T)
return clusters, W, H
Method 3: Similarity Network Fusion (SNF)
Integrate omics through patient similarity networks.
Phase 6: Pathway-Level Integration
Objective: Aggregate multi-omics evidence at the pathway level.
Approach: Score pathway dysregulation using combined evidence from multiple omics.
def integrate_pathway_evidence(omics_results, pathway_genes):
"""
Score pathway dysregulation across omics.
omics_results: {
'rnaseq': {'gene': fold_change},
'proteomics': {'gene': fold_change},
'methylation': {'gene': methylation_diff},
'cnv': {'gene': copy_number}
}
pathway_genes: List of genes in pathway
"""
# For each gene in pathway
pathway_scores = []
for gene in pathway_genes:
gene_score = 0
evidence_count = 0
# RNA-seq evidence
if gene in omics_results['rnaseq']:
gene_score += abs(omics_results['rnaseq'][gene])
evidence_count += 1
# Proteomics evidence
if gene in omics_results['proteomics']:
gene_score += abs(omics_results['proteomics'][gene])
evidence_count += 1
# Methylation evidence (negative correlation)
if gene in omics_results['methylation']:
gene_score += abs(omics_results['methylation'][gene])
evidence_count += 1
# CNV evidence
if gene in omics_results['cnv']:
gene_score += abs(omics_results['cnv'][gene])
evidence_count += 1
if evidence_count > 0:
pathway_scores.append(gene_score / evidence_count)
# Aggregate pathway score (mean of gene scores)
pathway_score = np.mean(pathway_scores) if pathway_scores else 0
return {
'pathway_score': pathway_score,
'n_genes_with_evidence': len(pathway_scores),
'n_omics_types': evidence_count
}
Use ToolUniverse enrichment tools:
# Get pathways for gene set
from tooluniverse import ToolUniverse
tu = ToolUniverse()
# Enrichment for genes dysregulated in ANY omics
all_dysregulated_genes = set()
all_dysregulated_genes.update(rnaseq_degs)
all_dysregulated_genes.update(diff_proteins)
all_dysregulated_genes.update(methylation_dmgs)
# Run enrichment
enrichment = tu.run_one_function({
"name": "enrichr_enrich",
"arguments": {
"gene_list": ",".join(all_dysregulated_genes),
"library": "KEGG_2021_Human"
}
})
# Score each pathway with multi-omics evidence
for pathway in enrichment['data']['results']:
pathway_genes = pathway['genes']
pathway['multi_omics_score'] = integrate_pathway_evidence(
omics_results, pathway_genes
)
Phase 7: Biomarker Discovery
Objective: Identify multi-omics signatures for disease classification.
Feature selection across omics:
def select_multiomics_features(X_dict, y, n_features=50):
"""
Select top features across omics for classification.
X_dict: {
'rnaseq': DataFrame (samples x genes),
'proteomics': DataFrame (samples x proteins),
'methylation': DataFrame (samples x CpGs)
}
y: Target labels (disease vs control)
Returns: Selected features per omics
"""
from sklearn.feature_selection import SelectKBest, f_classif
selected_features = {}
for omics_type, X in X_dict.items():
selector = SelectKBest(f_classif, k=min(n_features, X.shape[1]))
selector.fit(X, y)
# Get selected feature names
selected_idx = selector.get_support()
selected_features[omics_type] = X.columns[selected_idx].tolist()
return selected_features
Multi-omics classification:
def multiomics_classification(X_dict, y, selected_features):
"""
Train classifier using multi-omics features.
"""
from sklearn.ensemble import RandomForestClassifier
from sklearn.model_selection import cross_val_score
# Concatenate selected features from each omics
X_combined = []
for omics_type, features in selected_features.items():
X_combined.append(X_dict[omics_type][features])
X_combined = pd.concat(X_combined, axis=1)
# Train classifier
clf = RandomForestClassifier(n_estimators=100, random_state=42)
scores = cross_val_score(clf, X_combined, y, cv=5, scoring='roc_auc')
return {
'mean_auc': scores.mean(),
'std_auc': scores.std(),
'n_features': X_combined.shape[1],
'features_per_omics': {k: len(v) for k, v in selected_features.items()}
}
