🔬 CORE OMICS DISCIPLINE 03 LC-MS/MS • Discovery DIA/DDA • Phosphoproteomics • PTMs ⏱️ Comprehensive Production Blueprint • DIA-NN • MSstats • DEP

Quantitative Proteomics: Mass Spectrometry, Spectral Libraries & Phospho-Signaling Networks

Proteomics measures the ultimate functional effector molecules of the cell. Because post-transcriptional attenuation, spatial sequestering, and degradation decouple mRNA expression from actual protein abundance, high-resolution tandem mass spectrometry (LC-MS/MS) provides the indispensable direct readout of biological phenotypes, active post-translational modifications (PTMs), and druggable signaling pathways.

1. Biological Foundations 2. Mass Spectrometry Chemistries 3. Visual Architecture (SVG) 4. Computational Pipeline & Code 5. Real-World Case Study 6. Software Directory & Links 7. Pitfalls & QC Standards 8. Exercises & Datasets

🔬 Module 1: Biological Foundations & The Central Dogma Decoupling

Functional Biology

Proteins execute structural, catalytic, transport, and signaling functions. In classical molecular biology, mRNA was long assumed to be a direct proxy for protein copy numbers. Landmark proteogenomic studies have demonstrated that the cross-sample Spearman correlation between mRNA expression and protein abundance in human tissues averages only R ≈ 0.40.

Translational & Degradation Kinetics

Protein levels reflect the balance of translation rate (ribosome density and tRNA availability) and degradation rates governed by the ubiquitin-proteasome system and autophagy. Protein half-lives range from minutes (e.g. *MYC*, *p53*) to months (e.g. collagen, histones, nuclear pore complexes).

Post-Translational Modifications (PTMs)

PTMs dynamically rewire protein function without altering amino acid sequence:
• Phosphorylation (Ser/Thr/Tyr, +79.97 Da): Kinase/phosphatase switches regulating signaling.
• Ubiquitination (+114.04 Da diGly remnant): Targeting to 26S proteasome.
• Acetylation & Methylation: Chromatin histones and metabolic enzymes.

Protein Isoforms & Complexes

Proteins assemble into macromolecular complexes (e.g. spliceosome, ribosome, mediator complex). The physical interactome defines biological function rather than individual monomer abundance.

🔬 Module 2: Mass Spectrometry Architectures & Acquisition Modes

LC-MS/MS Instrumentation

Bottom-up proteomics digests intact proteins into tryptic peptides (cleaving C-terminal to Lysine and Arginine, except before Proline), separates them by hydrophobicity via nanoflow reverse-phase liquid chromatography (nanoLC), and ionizes them into the gas phase via electrospray ionization (ESI):

Acquisition Paradigm	Instrumental Mechanism	Quantification Basis	Missing Value Rate	Software Toolchain
DDA (Data-Dependent Acquisition)	MS1 survey scan selects TopN (e.g. 15–20) most abundant precursor ions for isolated fragmentation.	MS1 area under curve (MaxLFQ) or isobaric reporter ions (TMT 16-plex).	Moderate – High (Stochastic selection)	MSFragger, MaxQuant, Comet, SEQUEST
DIA (Data-Independent Acquisition)	Precursor mass range sliced into wide isolation windows (e.g. 32 × 25 m/z); all co-eluting ions fragmented simultaneously.	Fragment ion chromatographic peak extraction (MS2 level) using spectral libraries.	Ultra-Low (<5% across cohorts)	DIA-NN, Spectronaut, OpenSWATH
Targeted PRM / MRM	Quadrupole isolates pre-selected peptide precursor m/z; fragment ions continuously monitored across retention time window.	Absolute quantification using stable isotope-labeled standard (SIS) peptides.	0% (Deterministic monitoring)	Skyline, TargetQuant

🖼️ Module 3: Mass Spectrometry Physics & MS/MS Spectra

Scientific Diagram

²⁺, [M+3H]³⁺) Thermo Orbitrap Exploris / Astral Resolution: 120,000 • Accuracy: <2 ppm 2. Precursor MS1 Spectrum m/z (Mass-to-Charge) m/z 645.32 [M+2H]²⁺ Precursor Isolation Window: ±0.7 m/z (DDA) 3. MS2 Fragmentation (b & y Series) Fragment m/z b₂ b₄ y₅ y₇ Peptide: LVNEVTEFAK (Albumin)

Figure 3.1: Tandem Mass Spectrometry (LC-MS/MS) Quantification Principles — (1) Nano-flow reverse-phase gradient elution and electrospray ionization; (2) Precursor mass-to-charge survey scanning; (3) Collision-induced dissociation producing deterministic b-type and y-type peptide sequence ladders.

💻 Module 4: Computational Pipeline & Production Code

DIA-NN & R DEP

Step 1 & 2: Raw Centroiding & Library-Free DIA-NN CLI Execution

#!/usr/bin/env bash
set -euo pipefail

# 1. Convert vendor .raw files to indexed centroided .mzML
ThermoRawFileParser.sh -d ./raw_spectra/ -o ./mzml_centroided/ -f 2 -g -p

# 2. Step 2A: Generate in-silico deep-learning predicted spectral library from FASTA
diann \
  --lib "" \
  --threads 32 \
  --verbose 1 \
  --out-lib reference_insilico.speclib \
  --gen-spec-lib \
  --predictor \
  --fasta Homo_sapiens.GRCh38.pep.all.fa \
  --fasta-search \
  --min-fr-mz 200 --max-fr-mz 1800 \
  --min-pep-len 7 --max-pep-len 30 \
  --min-pr-mz 300 --max-pr-mz 1800 \
  --min-pr-charge 1 --max-pr-charge 4 \
  --cut K*,R* --missed-cleavages 1 \
  --unimod4 \
  --var-mods 1 --var-mod UniMod:35,15.994915,M

# 3. Step 2B: High-throughput quantification with Match-Between-Runs (MBR) & MaxLFQ
diann \
  --f ./mzml_centroided/*.mzML \
  --lib reference_insilico.speclib.predicted.speclib \
  --threads 32 \
  --out diann_report.tsv \
  --qvalue 0.01 \
  --matrices \
  --pg-level 1 \
  --reanalyse \
  --smart-profiling

Step 3 & 4: VSN Normalization, MinProb MNAR Imputation & Limma (R)

# Production R Script: Differential Enrichment of Proteomics (DEP) & Imputation
library(DEP)
library(dplyr)
library(SummarizedExperiment)

# 1. Ingest DIA-NN Protein Group matrix and clean identifiers
pg_raw <- read.delim("diann_report.pg_matrix.tsv", stringsAsFactors = FALSE, check.names = FALSE)
pg_clean <- pg_raw %>%
  filter(!grepl("cRAP_|CON__|REV__", Protein.Group)) %>%
  make_unique("Genes", "Protein.Group", delim = ";")

# 2. Define experimental design table
exp_design <- data.frame(
  label = c("Ctrl_1", "Ctrl_2", "Ctrl_3", "Treat_1", "Treat_2", "Treat_3"),
  condition = c("Control", "Control", "Control", "Treated", "Treated", "Treated"),
  replicate = c(1, 2, 3, 1, 2, 3)
)

# 3. Assemble SummarizedExperiment and filter missing values (require in >= 2/3 replicates)
sample_cols <- which(colnames(pg_clean) %in% exp_design$label)
se <- make_se(pg_clean, sample_cols, exp_design)
se_filt <- filter_missval(se, thr = 1)

# 4. Variance Stabilizing Normalization (VSN) & Left-Censored MNAR Imputation (MinProb)
se_norm <- normalize_vsn(se_filt)
se_imp <- impute(se_norm, fun = "MinProb", q = 0.01)

# 5. Moderated t-test differential testing via limma
se_diff <- test_diff(se_imp, type = "control", control = "Control")
dep_res <- add_rejections(se_diff, alpha = 0.05, lfc = log2(2.0))
write.csv(get_results(dep_res), "proteomics_differential_proteins.csv", row.names = FALSE)

🏥 Module 5: Real-World Case Study

Phosphoproteomic Oncology

Case: Uncovering MET Receptor Tyrosine Kinase Bypass in Osimertinib-Resistant NSCLC

Clinical Problem: Non-Small Cell Lung Cancer (NSCLC) patients with *EGFR*^T790M mutations receiving third-generation osimertinib frequently develop therapeutic resistance in the absence of secondary *EGFR* mutations (e.g. C797S). High-resolution DIA phosphoproteomics was performed on pre- and post-resistance patient-derived tumor models.

Quantified Coverage: 6,420 Protein Groups • 14,810 High-Confidence Class-I Phosphosites (Ascore > 19)

Top Phosphosite: MET Receptor Tyrosine Kinase (Y1234/Y1235, log2FC = +3.92, padj = 3.2e-14)

Downstream Substrates: GAB1 (Y627-p, log2FC = +3.10) • MAPK3/ERK1 (T202/Y204-p, log2FC = +2.78)

Kinase-Substrate Enrichment (KSEA): MET Kinase z-score = +4.12, p < 0.0001

Translational Resolution: Phosphoproteomic profiling proved that MET kinase bypass activation was actively driving downstream ERK signaling despite total EGFR shutdown. Combination treatment with osimertinib and capmatinib (MET inhibitor) induced rapid tumor regression.

📦 Module 6: Essential Software Directory & Official Links

Verified Tool Stack

DIA-NN

DIA Engine

Universal software for data-independent acquisition proteomics using deep neural networks and MaxLFQ.

Official DIA-NN GitHub →

FragPipe (MSFragger)

DDA / Open Search

Ultra-fast peptide identification engine by the Nesvizhskii lab capable of open modification mass searches.

FragPipe Suite Portal →

Skyline

Targeted MRM/PRM

Freely-available Windows client for building selected reaction monitoring (SRM/MRM) and PRM targeted assays.

MacCoss Lab Skyline →

MSstats

Statistical Modeling

R package for statistical relative quantification of proteins in mass spectrometry-based proteomics.

MSstats Official Site →

ThermoRawFileParser

Raw Conversion

Cross-platform C# command line utility for parsing Thermo .raw files to indexed .mzML on Linux/macOS.

CompOmics GitHub →

KSEAapp

Kinase Activity

Kinase-Substrate Enrichment Analysis tool inferring upstream kinase activity scores from phosphoproteomics data.

KSEAapp Package →

⚠️ Module 7: Bioinformatic Pitfalls & Quality Control Standards

Best Practices

1. The Protein Inference Ambiguity

Because multiple protein isoforms or paralogs share identical tryptic peptide sequences (shared peptides), attributing peptide signals to a single parent protein is ambiguous. Enforce Occam’s razor protein parsimony grouping.

2. MAR vs MNAR Imputation Bias

Missing values in proteomics are predominantly Missing Not At Random (MNAR)—caused by peptide abundances falling below instrument detection limits. Imputing with mean or kNN introduces severe false-positives; use left-censored Gaussian random draws (MinProb/QRILC).

3. False Discovery Rate (FDR) Propagation

A 1% peptide-spectrum match (PSM) FDR accumulates to ~5–10% FDR at the protein group level when tens of thousands of peptides are assembled. Always enforce 1% FDR at both the precursor and protein group levels.

📝 Module 8: Self-Assessment Exercises & Practice Datasets

Hands-on Mastery

📦 Curated Benchmark: DIA-NN Protein Group Matrix (proteomics_mini_diann.tsv)

Size: 4.2 MB • 600 Human Protein Groups across 6 LC-MS runs for DEP and limma practice.

Access Practice Dataset →

Q1: Why does Data-Independent Acquisition (DIA) generate multiplexed MS2 spectra, and how is it resolved?

Answer: In DIA, the quadrupole isolates wide precursor m/z windows (e.g. 25 m/z wide) containing dozens of co-eluting peptides, all of which enter the collision cell and fragment simultaneously. The resulting composite MS2 spectrum contains intertwined fragment ions. DIA-NN and Spectronaut resolve this complexity by matching the elution profiles of fragment ions against predicted or empirical spectral libraries using deep neural networks.

Q2: What is the primary difference between b-ions and y-ions in collision-induced dissociation (CID/HCD)?

Answer: When the peptide backbone amide bond (peptide bond) is fragmented by high-energy collisions, charge retention determines ion series: if the positive charge is retained on the N-terminal fragment, a b-ion is formed; if the charge is retained on the C-terminal fragment, a y-ion is formed. Because tryptic peptides end in basic Lysine or Arginine, y-ions generally exhibit higher intensity and cleaner sequence ladders.

Q3: Why is Variance Stabilizing Normalization (VSN) preferred over simple log2 transformation in proteomics?

Answer: At low mass spectrometry peak intensities, measurement variance is dominated by additive instrument noise, whereas at high intensities, variance is multiplicative. A standard log transformation over-inflates the variance of low-intensity peptides near the detection limit. VSN applies an inverse hyperbolic sine transformation ($arsinh$) that stabilizes variance uniformly across the entire dynamic range.