omics-integration

Multi-Omics Integration

Overview

Multi-omics integration combines data from multiple molecular measurement platforms (RNA-seq, proteomics, ATAC-seq, metabolomics, methylation) to identify shared biological variation and cross-modal regulatory relationships. This skill covers MOFA+ for unsupervised factor analysis, MixOmics/DIABLO for supervised integration, and network-based approaches.

When to Use This Skill

• Integrating paired multi-omics datasets (same samples across platforms)
• Identifying latent factors that explain variation across omics layers
• Building multi-omics classifiers or biomarker panels
• Finding correlated features across genomics, transcriptomics, and proteomics
• Studying regulatory cascades (DNA → RNA → protein → metabolite)
• Visualizing multi-omics data in low-dimensional space
• Performing network-based multi-omics enrichment

Quick Start

MOFA+ (Multi-Omics Factor Analysis)

from mofapy2.run.entry_point import entry_point
import pandas as pd
import numpy as np

# Prepare multi-omics data as list of DataFrames
# Each DataFrame: samples × features, one per omics layer
rna_data = pd.read_csv("rna_normalized.csv", index_col=0)      # 100 samples × 20000 genes
protein_data = pd.read_csv("protein_lfq.csv", index_col=0)     # 100 samples × 5000 proteins
metabolite_data = pd.read_csv("metabolites.csv", index_col=0)  # 100 samples × 1000 metabolites

# Align samples
common_samples = rna_data.index.intersection(protein_data.index).intersection(metabolite_data.index)
rna_data = rna_data.loc[common_samples]
protein_data = protein_data.loc[common_samples]
metabolite_data = metabolite_data.loc[common_samples]

print(f"Shared samples: {len(common_samples)}")
print(f"RNA: {rna_data.shape[1]} features")
print(f"Protein: {protein_data.shape[1]} features")
print(f"Metabolite: {metabolite_data.shape[1]} features")

# Prepare MOFA+ input
ent = entry_point()
ent.set_data_options(scale_groups=False, scale_views=False)
ent.set_data_df(
    pd.concat([rna_data.T, protein_data.T, metabolite_data.T]),
    likelihoods=["gaussian", "gaussian", "gaussian"],
)
ent.set_model_options(factors=15, spikeslab_weights=True, ard_factors=True, ard_weights=True)
ent.set_train_options(iter=1000, convergence_mode="fast", seed=42, gpu_mode=False)

ent.build()
ent.run()
ent.save("mofa_model.hdf5")
print("MOFA+ training complete.")