proteomics

Proteomics Analysis

Method selection and execution guidance for quantitative proteomics across mass spectrometry and affinity-based platforms.

Platform Detection and Preprocessing

Identify the platform first — each has distinct data characteristics and preprocessing requirements:

Platform?
├── DDA (Data-Dependent Acquisition)
│   ├── Input: FragPipe/MaxQuant output → protein/peptide intensity matrix
│   ├── Preprocessing:
│   │   ├── 1. Filter by missingness (keep proteins detected in >= 70% of at least one group)
│   │   ├── 2. Log2 transform intensities
│   │   ├── 3. Normalize: VSN (variance stabilizing, preferred) or median centering
│   │   ├── 4. Impute missing values:
│   │   │   ├── MNAR (Missing Not At Random, below LOD) → MinProb (left-censored)
│   │   │   ├── MAR (Missing At Random, technical dropout) → kNN
│   │   │   └── Mixed → hybrid (MinProb for low-abundance, kNN for rest)
│   │   └── 5. Assess batch effects (PCA by batch)
│   └── DE: DEP test_diff (default) or limma via rpy2
│
├── DIA (Data-Independent Acquisition)
│   ├── Input: DIA-NN output → protein/precursor intensity matrix
│   ├── Preprocessing: Same as DDA but fewer missing values expected
│   │   ├── DIA-NN already provides normalized intensities (check MaxLFQ column)
│   │   └── If raw intensities: filter → log2 → normalize → impute as above
│   └── DE: DEP test_diff or limma via rpy2
│
├── TMT / iTRAQ (Isobaric Labeling)
│   ├── Input: FragPipe/MaxQuant PSM-level or protein-level quantification
│   ├── Preprocessing:
│   │   ├── 1. MSstatsTMT dataProcess (handles plex normalization, protein summarization)
│   │   ├── 2. Channel-level normalization (reference channel, global median)
│   │   └── 3. Between-plex normalization (bridge channel if available)
│   └── DE: MSstatsTMT groupComparison (handles plex as random effect)
│
├── Olink (Proximity Extension Assay)
│   ├── Input: NPX values (already log2-scale, plate-normalized)
│   ├── Preprocessing:
│   │   ├── 1. LOD filtering: flag assays below LOD per sample
│   │   ├── 2. Multi-plate: bridge normalization using bridge samples
│   │   ├── 3. QC warnings: remove samples/assays with QC_Warning flags
│   │   └── 4. No additional normalization needed (NPX is pre-normalized)
│   └── DE: limma via rpy2 (treats NPX as log2-intensity) or linear model
│
└── SomaScan (SOMAmer Aptamer)
    ├── Input: ADAT file → RFU (Relative Fluorescence Units) matrix
    ├── Preprocessing:
    │   ├── 1. ANML normalization (SomaScan standard, usually pre-applied)
    │   ├── 2. Log2 transform RFU values
    │   ├── 3. Filter: remove aptamers with high CV across controls
    │   └── 4. Plate/batch correction if multi-plate
    └── DE: limma via rpy2 on log2 RFU values