What Proteomics Is Used For

Proteomics is used for four things that matter in practice. It finds biomarkers that track a disease or a physiological state and it identifies and validates drug targets. It also measures what a drug does to a body after it is given, and it builds a personal baseline against which later measurements can be compared. Everything else in the field, including structural work and interaction mapping, feeds one of those four. The reason proteins earn this attention is that they are the layer where genotype, transcription, and environment all resolve into something that does work in a cell. A genome tells you what is possible. A proteome tells you what is present right now, at what concentration, and in what modified form.

This page covers the main types of proteomics and what each one is good for. It also describes what the data files look like and how we would analyze a longitudinal plasma proteome from one person.

The main types of proteomics

Expression proteomics, sometimes called discovery or shotgun proteomics, asks which proteins are present in a sample and at what abundance. This is the workhorse. It is done by mass spectrometry after the proteins are digested into peptides with trypsin, and it produces a quantitative matrix of proteins by samples.

Targeted proteomics inverts the question. You choose a set of proteins in advance and measure them with high precision using selected or parallel reaction monitoring (SRM and PRM). Fewer analytes, far better reproducibility, and the ability to add stable isotope labeled standards for absolute quantification in nanograms per milliliter.

Affinity proteomics uses binders instead of a mass spectrometer. Olink uses paired antibodies and a proximity extension assay with a DNA readout. SomaScan uses modified aptamers. Both reach several thousand plasma proteins with small sample volumes, which is why most large population cohorts use them.

Post-translational modification proteomics measures the chemistry layered on top of the sequence. Phosphoproteomics, done with titanium dioxide or Fe-IMAC enrichment, is the standard readout for signaling. Glycoproteomics and ubiquitinomics follow the same pattern with different enrichment chemistry. Interaction proteomics (affinity purification followed by mass spectrometry, or proximity labeling with BioID and TurboID) maps which proteins are physically near each other. Structural proteomics resolves shapes, and now leans on prediction as much as on crystallography ¹.

Biomarker discovery, and why most candidates fail

Biomarker discovery is the most cited application and the one with the worst hit rate. The cancer literature illustrates the pattern well: hundreds of candidate markers have been reported from tissue and plasma studies, and very few have survived independent validation into clinical use ². The usual causes are sample handling differences between cases and controls, batch structure that aligns with group membership, and an effect size that shrinks once the cohort is not the discovery cohort.

The dynamic range of plasma is the physical reason discovery is hard. Protein concentrations span roughly ten orders of magnitude, and a handful of proteins led by albumin account for most of the mass. Undepleted plasma on a modern instrument gets you a few hundred proteins. Depletion of the top 14 abundant proteins, or nanoparticle-based enrichment, pushes that into the low thousands. The markers people care about usually sit below that line.

This is a known bottleneck for translating omics into clinical practice, alongside standardization and cost ³. It is also why we treat any single-timepoint plasma result as a hypothesis rather than a finding.

Drug discovery and pharmacodynamics

In drug development, proteomics is applied at three points. Target identification uses expression and interaction data to decide what to drug. Mechanism work uses phosphoproteomics and thermal proteome profiling to check that a compound engages the intended protein and to find what else it hits. Safety and pharmacodynamics use plasma and tissue proteomes to detect organ stress earlier than standard chemistry panels do ⁴.

The integration with genomics and metabolomics is what makes this useful rather than descriptive. Genomic evidence tells you a target is causal. Proteomic evidence tells you whether the pathway moved in a treated animal or a treated person ⁵. Both are needed before a program is worth its cost.

Precision medicine and the personal baseline

The application that matters most to an individual reader is the least discussed in review articles. Population studies compare groups. A personal proteome compares you to yourself over time, which removes the between-person variation that dominates most plasma protein measurements.

The argument for this in clinical proteomics has been made carefully by Van Eyk and Snyder, who frame the individual’s own profile as the reference against which change is judged ⁶. For many plasma proteins, within-person variation over months is much smaller than the spread across a population. A value inside the population reference interval can still be a large personal deviation.

This is where proteomics complements genomics and transcriptomics rather than repeating them ⁷. Your genome does not change. Your transcriptome shifts on a timescale of hours. Your plasma proteome moves on a timescale of days to weeks, which is close to the timescale of most interventions a person would try.

The files and the tools

If you get mass spectrometry data, you get vendor raw files (Thermo .raw, Bruker .d) and should immediately convert to open format with msconvert --mzML --filter "peakPicking vendor msLevel=1-2". Keep the raw files. Search engines improve, and you will want to reprocess.

For data-dependent acquisition, we use FragPipe with MSFragger. A closed search with 20 ppm precursor tolerance and 1% false discovery rate at both peptide and protein level is the sane default. For data-independent acquisition, which is what most modern plasma work uses, our recommendation is DIA-NN in library-free mode. Point it at a FASTA file and let it predict a spectral library. Then enable match-between-runs across your timepoints. Output is a long-format report.tsv plus a protein group matrix. Skyline is the tool for targeted PRM work and for inspecting individual peptide chromatograms when a result looks strange.

Affinity platforms deliver tables rather than spectra. Olink gives NPX values, which are on a log2 scale and normalized within each assay. NPX is relative, so a value of 4 for one protein and 4 for another says nothing about their concentrations. SomaScan gives an .adat file with relative fluorescence units and its own normalization steps already applied.

Three failure modes cause most bad analyses. Missing values in mass spectrometry data are mostly left-censored, meaning the protein was below detection rather than absent, so mean imputation is wrong and a low-value imputation such as QRILC is closer to the truth. Affinity data carries explicit limit-of-detection flags, and proteins with most samples below LOD should be dropped rather than modeled. Batch and plate effects are large enough to create clean-looking false results, so record which plate and which run each sample came from and include that term in the model.

For longitudinal analysis in R, we fit lmer(NPX ~ time + plate + (1 | protein)) per protein. More usefully for one person, we plot each protein’s trajectory against its own measured variability. Correct across proteins with Benjamini-Hochberg. With a single subject you do not get a p-value in the usual sense. You get a trajectory, and the useful question is whether a change exceeds the assay’s coefficient of variation and your own historical range.

Where you need a clinician

Proteomic panels are measurement tools. Almost none of the several thousand analytes on a research platform have a validated clinical decision threshold. The ones that do, namely troponin and PSA and CRP, are run on regulated clinical assays rather than research panels. A value that looks alarming on a research platform should be re-measured on a clinical assay before anyone draws a conclusion, and that conversation belongs with a physician. This gap between research-grade and clinical-grade measurement is the central obstacle to putting proteomics into routine care ³.

Questions people also ask

What are examples of proteomics?

Examples include measuring the phosphorylation state of a kinase pathway after a drug is added to cells and profiling several thousand plasma proteins across a cohort to find markers of a disease. Others are mapping which proteins bind a transcription factor by affinity purification and mass spectrometry, and quantifying a specific protein in serum by parallel reaction monitoring with an isotope-labeled standard.

Who is the father of proteomics?

Marc Wilkins is usually credited with coining the term “proteome” in 1994 while a graduate student working with Keith Williams in Australia. The techniques predate the word by a decade or more, and two-dimensional gel electrophoresis from the 1970s is the direct ancestor of the field.

What are the clinical applications of proteomics?

Established clinical uses are narrow and targeted: cardiac troponin for myocardial injury, serum protein electrophoresis for monoclonal gammopathy, and mass spectrometry typing of amyloid deposits. Broader applications in patient stratification and treatment monitoring are advancing but remain largely research-stage ⁶.

Is proteomics the future?

It is one necessary layer among several. Proteomics resolves questions that genome sequences cannot answer, because it measures what is present rather than what is encoded. The strongest results come from combining it with genomic, transcriptomic and metabolomic data on the same samples ^{8 9}.

What are the seven main types of proteins?

The textbook classification lists antibodies and contractile proteins and enzymes. It also lists hormonal proteins and structural proteins, along with storage proteins and transport proteins. This is a functional teaching category rather than anything used in proteomics analysis, where proteins are grouped by pathway, localization or complex membership.

Footnotes

1. John R. Yates, Andrei L. Osterman. Introduction: Advances in Genomics and Proteomics. Chemical Reviews, 2007. https://doi.org/10.1021/cr068201u ↩

2. Yang Woo Kwon, Han-Seul Jo, Sungwon Bae, et al. Application of Proteomics in Cancer: Recent Trends and Approaches for Biomarkers Discovery. Frontiers in Medicine, 2021. https://doi.org/10.3389/fmed.2021.747333 ↩

3. Edouard C. Nice. Challenges for omics technologies in the implementation of personalized medicine. Expert Review of Precision Medicine and Drug Development, 2018. https://doi.org/10.1080/23808993.2018.1505429 ↩ ↩²

4. Mollisa M. Elrick, Jennie L. Walgren, Michael D. Mitchell, et al. Proteomics: Recent Applications and New Technologies. Basic & Clinical Pharmacology & Toxicology, 2006. https://doi.org/10.1111/j.1742-7843.2006.pto_391.x ↩

5. Craig Russell, Ayesha Rahman, Afzal R Mohammed. Application of genomics, Proteomics and Metabolomics in Drug discovery, Development and Clinic. Therapeutic Delivery, 2013. https://doi.org/10.4155/tde.13.4 ↩

6. Jennifer E. Van Eyk, Michael P. Snyder. Precision Medicine: Role of Proteomics in Changing Clinical Management and Care. Journal of Proteome Research, 2018. https://doi.org/10.1021/acs.jproteome.8b00504 ↩ ↩²

7. Richard P Horgan, Louise C Kenny. ‘Omic’ technologies: genomics, transcriptomics, proteomics and metabolomics. The Obstetrician & Gynaecologist, 2011. https://doi.org/10.1576/toag.13.3.189.27672 ↩

8. A. E. Brunetti, F. Carnevale Neto, M. C. Vera, et al. An integrative omics perspective for the analysis of chemical signals in ecological interactions. Chemical Society Reviews, 2018. https://doi.org/10.1039/c7cs00368d ↩

9. Dr. Seyi Enitan, Esther Adejumo, Dr. Osareniro Osakue, et al. Integrating Genomics and Proteomics Technologies in Biological Research: Advantages, Challenges, and Prospects. Global South Health Horizons, 2025. https://doi.org/10.63950/gshh.2025.1.1.2 ↩