How to Process Shotgun Proteomics Data in Multi-Omics Integration?

In multi-omics integration analyses, appropriate processing of Shotgun proteomics data is of critical importance. Proteomics data are inherently semi-quantitative, characterized by a high proportion of missing values and substantial variability across samples and analytical platforms. Inadequate data handling can therefore substantially impair integrative analyses with other omics layers, including genomics, transcriptomics, and metabolomics.

Shotgun Proteomics: Data Characteristics and Challenges

Shotgun proteomics, also referred to as data-dependent acquisition (DDA) proteomics, is a mass spectrometry–based approach for comprehensive profiling of complex protein mixtures. The primary output typically consists of a protein relative abundance matrix. Its major characteristics include:

• Relative quantification, such as LFQ (Label-Free Quantification) or iBAQ (Intensity-Based Absolute Quantification)
• A high proportion of missing values, with certain proteins undetected in specific samples
• Strongly skewed intensity distributions, commonly approximating a log-normal distribution
• Hierarchical data organization spanning peptides, proteins, and pathways
• Limited direct correspondence across omics layers, for example, between protein identifiers, transcript IDs, and metabolite names

Consequently, rigorous preprocessing of Shotgun proteomics data constitutes a fundamental prerequisite for high-quality multi-omics integration.

Proteomics Data Preprocessing

1. Data Filtering

(1) Removal of proteins with low identification confidence (e.g., PEP > 0.01)

(2) Exclusion of contaminants and reverse-sequence identifications

(3) Elimination of proteins with extremely low detection frequency (e.g., missing in more than 50% of samples)

2. Logarithmic Transformation

Log2 transformation is commonly applied to reduce distributional skewness and approximate normality, thereby enhancing the robustness of downstream statistical analyses.

3. Batch Effect Correction

When integrating data across multiple experimental batches or platforms, batch effects are typically corrected using ComBat (implemented in the sva package) to improve cross-sample comparability.

Missing Value Handling: Challenge and Biological Signal

Missing values represent a defining feature of Shotgun proteomics data, and their treatment can profoundly influence downstream analytical outcomes. It is essential to distinguish between different missingness mechanisms:

1. Missing at Random (MAR)

Such missingness may result from stochastic effects or instrument sensitivity limitations. Methods such as k-nearest neighbors (KNN) imputation and Bayesian principal component analysis are commonly applied.

2. Missing Not at Random (MNAR)

This type of missingness is frequently associated with low-abundance proteins falling below detection limits and may carry biological relevance. Left-censored imputation approaches, including MinProb and QRILC, are therefore recommended.

Protein Annotation and Gene Mapping: Enabling Cross-Omics Alignment

A central requirement for multi-omics integration is the establishment of a shared reference framework, enabling alignment between Shotgun proteomics data and other omics layers such as transcriptomics, genomics, and metabolomics. Common practices include:

• Mapping UniProt identifiers or accession numbers to gene symbols or Ensembl IDs
• Performing batch annotation conversion using tools such as bioMart, the UniProt API, and gProfiler
• Consolidating peptide-level identifications into unique representative proteins through protein inference
• Standardizing nomenclature to resolve inconsistencies across databases and analytical platforms

Dimensionality Reduction and Feature Selection: Laying the Foundation for Integration

Prior to integrative analysis, dimensionality reduction and feature selection are indispensable steps for Shotgun proteomics data:

• Variance-based filtering to retain proteins exhibiting substantial inter-sample variability
• PCA, t-SNE, or UMAP to explore sample relationships and clustering patterns
• WGCNA to construct weighted co-expression networks and derive module-level features
• Differential analysis to identify and retain significantly differentially expressed proteins (DEPs) for subsequent modeling or enrichment analyses

Collectively, these approaches improve data interpretability and facilitate effective integration with other omics layers.

Strategies for Integration with Other Omics Data

Depending on study objectives and data characteristics, several integration strategies are commonly employed:

1. Vertical Integration

Vertical integration refers to cross-layer analysis (e.g., mRNA, protein, and metabolite data), emphasizing coordinated regulation of shared pathways or biological functions across omics levels.Representative methods include Multi-Omics Factor Analysis (MOFA), DIABLO (from mixOmics), iClusterPlus, and JIVE.

2. Horizontal Integration

Horizontal integration focuses on combining data within the same omics layer (e.g., multiple Shotgun proteomics batches), with an emphasis on consistent sample-level patterns.Common approaches include ComBat-based batch correction, Harmony, and canonical correlation analysis (CCA).

3. Pathway- or Network-Driven Integration

Pathway- or network-based strategies leverage prior knowledge from databases such as KEGG, Reactome, and STRING to identify pathways or network modules that are consistently enriched across multiple omics layers.

Multi-omics research is progressively evolving from simple data juxtaposition toward true information-level integration. Rigorous processing of Shotgun proteomics data not only determines the quality of integrative analyses but also directly influences result interpretability and translational potential. At MtoZ Biolabs, integrated mass spectrometry platforms, standardized data-processing workflows, and intelligent integrative algorithm pipelines are employed to support researchers in extracting biologically meaningful insights from complex omics datasets and advancing biomedical research.

MtoZ Biolabs, an integrated chromatography and mass spectrometry (MS) services provider.

Related Services

Protein Identification Service by Shotgun Proteomics