Proteomics

Agents

proteomics_expert

Parent orchestrator agent for proteomics analysis. Handles QC, preprocessing, and normalization directly, while delegating downstream analysis to specialized sub-agents.

Capabilities:

• MS data loading and preprocessing
• Protein quantification (LFQ, TMT, iTRAQ)
• Platform auto-detection (mass spectrometry vs affinity)
• Quality control with platform-specific metrics
• Normalization (median/log2 for MS, quantile for affinity)
• Pattern analysis with PCA and clustering
• Delegates DE analysis to proteomics_de_analysis_expert
• Delegates biomarker discovery to biomarker_discovery_expert

proteomics_de_analysis_expert

Sub-agent for differential expression analysis, invoked by proteomics_expert.

Capabilities:

• Differential protein expression between groups (7 statistical methods)
• Time course analysis for longitudinal studies
• Protein-target correlation analysis (Pearson, Spearman, Kendall)
• Platform-aware fold change thresholds (1.5x MS, 1.2x affinity)
• Multiple testing correction (Benjamini-Hochberg, Bonferroni)

biomarker_discovery_expert

Sub-agent for network analysis and survival-based biomarker identification, invoked by proteomics_expert.

Capabilities:

• WGCNA-style co-expression network analysis
• Module eigengene computation and trait correlation
• Cox proportional hazards regression
• Kaplan-Meier survival analysis with log-rank tests
• Biomarker candidate ranking

Example Workflows

Differential Protein Analysis

User: Analyze my proteomics data and find differentially
      abundant proteins between conditions

[proteomics_expert → proteomics_de_analysis_expert]
- Loads MS quantification data (MaxQuant, DIA-NN)
- Normalizes protein abundances
- Delegates to proteomics_de_analysis_expert for differential analysis
- Generates volcano plot
- Reports top differentially abundant proteins

PTM Analysis

User: Identify phosphorylation changes in my dataset

[proteomics_expert]
- Filters for phosphopeptides
- Normalizes phosphosite intensities
- Calculates differential phosphorylation
- Maps sites to protein domains
- Reports kinase motif enrichment

Protein Identification

User: What proteins were identified in my MS run?

[proteomics_expert]
- Summarizes identified proteins
- Reports coverage and peptide counts
- Shows intensity distribution
- Highlights high-confidence identifications

Multi-Omics Integration

proteomics_expert can integrate with transcriptomics data:

User: Compare protein and mRNA levels for my samples

[proteomics_expert + transcriptomics_expert]
- Matches proteins to transcripts
- Correlates protein/mRNA abundances
- Identifies discordant genes
- Reports post-transcriptional regulation candidates

Integration Workflow

1. Load RNA-seq data (transcriptomics_expert)
2. Load proteomics data (proteomics_expert)
3. Match gene IDs between datasets
4. Calculate correlation (Pearson/Spearman)
5. Identify outliers (high mRNA, low protein)
6. Annotate with pathway enrichment

Supported Data Formats

Services

lobster-proteomics includes domain-specific services bundled with the package:

Data Parsers

Services and parsers are installed automatically with the agent package.

Dependencies

lobster-proteomics requires proteomics analysis libraries:

These are installed automatically with the package.

Sub-Agent Architecture

proteomics_expert (supervisor-accessible)
├── proteomics_de_analysis_expert (sub-agent, not directly accessible)
└── biomarker_discovery_expert (sub-agent, not directly accessible)

The proteomics_expert handles QC, preprocessing, normalization, and pattern analysis directly. When the user requests differential expression, time course, or correlation analysis, it delegates to proteomics_de_analysis_expert. For network analysis (WGCNA) and survival analysis, it delegates to biomarker_discovery_expert.

Configuration

# .lobster_workspace/config.toml
enabled = ["proteomics_expert"]

Access

lobster-proteomics is free and open source. Install and use without any license or activation.

Quantification Methods