🔬 High-Resolution LC-MS/MS Data Science
Modern bottom-up mass spectrometry identifies thousands of proteins in complex biological lysates. MaxQuant and MSFragger provide state-of-the-art algorithms for high mass accuracy peptide-spectrum matching (PSM), label-free quantification (MaxLFQ), isobaric multiplexing (TMT), and post-translational modification (PTM) localization.
🔬 Peptide Monoisotopic Mass & Fragment Ion Ladder
Interactive ToolPeptide Sequence
Precursor m/z by Charge State
⚖️ Proteomics Quantification Method Advisor
Method SelectorWorkflow Preference
Key Characteristics
0. Bottom-Up Mass Spectrometry Physics
In shotgun bottom-up proteomics, complex protein extracts are enzymatically digested into peptides, separated by nano-LC reversed-phase chromatography, electrospray ionized (ESI), and analyzed in a high-resolution mass spectrometer (e.g. Orbitrap, timsTOF). Survey MS1 scans determine peptide precursor $m/z$ and intensity, while MS2 tandem fragmentation spectra provide sequence information.
1. Vendor RAW to Open mzML Conversion (msconvert)
Convert proprietary mass spec vendor files (Thermo .raw, Sciex .wiff, Bruker .d) into open standard mzML using ProteoWizard with 64-bit centroiding.
# Centroid raw MS1 and MS2 scans to mzML msconvert sample.raw \ --mzML --filter "peakPicking true 1-" \ --filter "zeroSamples removeExtra" \ -o mzml_output/
2. Target-Decoy FASTA & Contaminants Database (cRAP)
Concatenate canonical protein sequences (UniProt Swiss-Prot) with reversed decoy sequences and the common Repository of Adventitious Proteins (cRAP - keratins, trypsin, bovine serum).
# Combine UniProt human proteome with common contaminants cat UP000005640_human_sprot.fasta crap_contaminants.fasta > reference_proteome.fasta
3. Protease Digestion Parameters
Standard digestion utilizes Trypsin (cleaves C-terminal to Arg and Lys, blocked by Proline: Trypsin/P). Permit up to 2 missed cleavages to account for adjacent basic residues.
# Typical search engine enzyme configuration enzyme = Trypsin/P max_missed_cleavages = 2 min_peptide_len = 7 max_peptide_len = 45
4. Fixed & Variable PTM Modification Rules
Specify fixed modifications from sample alkylation (+57.021 Da Carbamidomethyl on Cysteine) and biological/oxidation variable modifications.
# Modifications Configuration Fixed: Carbamidomethyl [C] (+57.021464 Da) Variable: Oxidation [M] (+15.994915 Da) Variable: Acetyl [Protein N-term] (+42.010565 Da) Optional (Phosphoproteomics): Phospho [STY] (+79.966331 Da)
5. Peptide-Spectrum Matching (PSM) & Target-Decoy FDR
The search engine calculates theoretical fragment spectra and computes cross-correlation scores against acquired MS2 scans. A target-decoy strategy enforces a strict 1% False Discovery Rate at both PSM and protein levels.
6. MaxLFQ Normalization & Match-Between-Runs (MBR)
MaxLFQ extracts pairwise peptide ratio graphs to compute accurate relative protein abundances. Match-Between-Runs transfers peptide identifications from adjacent runs using calibrated retention time windows.
# MaxQuant CLI execution (configured via mqpar.xml) maxquant mqpar.xml
7. Ultrafast Open Search with MSFragger
MSFragger uses fragment-index matching to perform wide precursor tolerance open searches ([-500 Da, +500 Da]) 100x faster than traditional search engines, revealing unanticipated post-translational modifications.
# Execute MSFragger search with open mass window java -Xmx32G -jar MSFragger.jar \ closed_fragger.params \ sample_01.mzML sample_02.mzML
8. Differential Protein Expression in R (DEP / Limma)
Filter out contaminant proteins, log2-transform MaxLFQ intensities, normalize median intensity, and test differential abundance using empirical Bayes moderated t-tests.
library(DEP)
library(limma)
# Ingest MaxQuant proteinGroups.txt
data <- read.delim("proteinGroups.txt", stringsAsFactors = FALSE)
data_filtered <- data[data$Reverse != "+" & data$Potential.contaminant != "+", ]
9. Missing Value Imputation (Missing Not At Random - MNAR)
Mass spectrometry suffers from left-censored missing values (peptides below the limit of detection). Use left-shifted Gaussian truncation (QRILC or MinProb) to impute missing values realistically.
10. STRING-db Protein Interaction Networks & Pathways
Map significantly altered proteins onto physical protein-protein interaction networks and identify functional complexes.
library(STRINGdb) string_db <- STRINGdb$new(version="12.0", species=9606, score_threshold=400) mapped_proteins <- string_db$map(diff_proteins, "Gene.names", removeUnmappedRows = TRUE) string_db$plot_network(mapped_proteins$STRING_id)
🔀 Decision Architecture: Proteomics Technology Choice
Decision TreeCost-effective, zero reagent tagging cost, scales to hundreds of samples. Use MaxQuant or MSFragger FragPipe with Match-Between-Runs enabled to maximize identification transfer.