Proteomics measures the full complement of proteins in a biological sample using mass spectrometry (MS). In a typical bottom-up workflow, proteins are digested into peptides, separated by liquid chromatography, and analyzed by tandem mass spectrometry (LC-MS/MS). Computational analysis matches observed spectra to theoretical spectra from protein databases to identify peptides, then infers protein identities and quantities. Label-free quantification compares peptide intensities across runs, while labeling approaches (TMT, SILAC) enable multiplexed comparison. Proteomics captures information that transcriptomics cannot: protein abundance, post-translational modifications, protein-protein interactions, and protein turnover.
Analyze a published proteomics dataset using MaxQuant: load raw MS files, search against a protein database, filter by false discovery rate, and examine the identified proteins and their quantification. Compare the protein abundance rankings to RNA-seq expression data from the same tissue and observe the imperfect correlation.
Genomics tells you what genes an organism has. Transcriptomics tells you which genes are being transcribed. Proteomics tells you which proteins are actually present, at what levels, and in what modified forms — and since proteins are the primary functional molecules in cells, this is often the most biologically relevant layer of information.
The dominant technology is liquid chromatography-tandem mass spectrometry (LC-MS/MS). In the bottom-up workflow, proteins are extracted from a sample and digested into peptides using trypsin (which cuts at lysine and arginine residues). The peptide mixture is separated by liquid chromatography (typically reversed-phase HPLC), which reduces complexity by spreading peptides out over time. As peptides elute from the column, they are ionized (electrospray ionization) and enter the mass spectrometer, which measures their mass-to-charge ratio. Selected peptides are then fragmented (by collision with gas molecules), and the fragment masses are recorded. This fragmentation pattern is the peptide's "fingerprint" — it encodes the amino acid sequence.
Peptide identification matches these experimental fragmentation spectra to a database. For each spectrum, the search engine generates theoretical fragment spectra for all peptides in the database within the mass tolerance of the observed precursor, scores each match, and reports the best. This is a massive search problem — a human proteome database contains hundreds of thousands of possible peptide sequences. Statistical evaluation using the target-decoy approach ensures that the reported identifications have a controlled false discovery rate. Protein inference then groups identified peptides into protein groups, handling the complication that some peptides are shared between multiple proteins (the protein inference problem).
Quantification measures how much of each protein is present. Label-free quantification compares the intensity or spectral count of each peptide across runs, but requires careful normalization for run-to-run variability. Labeling approaches tag peptides from different conditions with different mass labels: TMT (tandem mass tags) allows up to 18 samples to be multiplexed in a single run, and SILAC (stable isotope labeling) incorporates heavy amino acids during cell growth for in vivo comparison. Each approach has tradeoffs in throughput, accuracy, and dynamic range. Beyond abundance, proteomics can map post-translational modifications (phosphorylation, ubiquitination, acetylation) that regulate protein activity, identify protein-protein interactions (co-immunoprecipitation MS), and measure protein turnover rates (pulsed SILAC) — information layers that no other technology provides.