Functional genomics and proteomics: how the layers differ and what you can do with your own
Proteomics does not fall under genomics. Genomics is the study of the DNA sequence itself, and proteomics is the measurement of the proteins present in a sample. Functional genomics is the larger question both serve: what does this genome do in this tissue, in this person, right now. The genome is a static parts list of roughly 20,000 protein-coding genes. The proteome is what the cell has built from that list this morning. Functional genomics is the set of methods used to get from the first to the second, and transcriptomics and proteomics are its two main experimental readouts 1.
For someone holding their own data, the distinction is not academic. Each layer answers a different class of question, fails in a different way, and arrives in a different file format. Knowing which layer can answer your question saves you from over-reading a variant that has no consequence, or from chasing a protein change that has a mundane technical explanation.
What each layer can and cannot tell you
Your whole genome is fixed. Sequenced at 30x coverage, it yields roughly four to five million positions where you differ from the reference, the large majority of which are common and unremarkable. The genome tells you about capacity and risk. It cannot tell you whether a gene is on.
The gap between sequence and function was the central problem left over from the Human Genome Project, and it was recognized as such from the start 2. Having the sequence created the annotation problem: deciding which stretches are genes, where transcripts begin and end, and what the products do 3. Comparative genomics was one of the first tools used to close that gap, since sequence that stays conserved across mammals is usually sequence that matters 4.
Transcriptomics is the first functional layer. RNA sequencing counts transcripts, so it tells you which genes are being read in the tissue you sampled, and at what relative abundance. It is sensitive, cheap per gene, and covers the whole transcriptome at once.
Proteomics is the layer closest to phenotype. Proteins do the work. Their abundance is set by transcription, translation rate, and degradation together. Because of that, protein levels are only loosely predicted by mRNA levels for many genes. Proteomics also sees phosphorylation, glycosylation, proteolytic cleavage, and complex formation, which the transcriptome cannot. Modifications of this kind are often the actual regulatory event, which is why proteomics is treated as the functional genomics readout that reports on state rather than potential 5.
The historical framing still holds up. Once genomes became routine, the bottleneck moved to the proteome, because protein abundance spans a far wider dynamic range than nucleic acids and no amplification step exists 6. In plasma the spread between albumin and a low-abundance cytokine is about ten orders of magnitude.
The three main types of proteomics
The field is conventionally split three ways, and the split matters because each type needs a different assay.
Expression proteomics asks which proteins are present and in what amount. This is what a plasma panel or a discovery mass-spectrometry run gives you. Structural proteomics asks where proteins sit and what they bind. It covers complexes, localization, and interaction maps. Functional proteomics asks what proteins are doing, which in practice means activity assays and post-translational modification mapping, most commonly phosphoproteomics.
For an individual working from blood, expression proteomics is the accessible one. The other two generally require cells, enrichment chemistry, or instruments you will not have. Integrating expression proteomics with genotype and transcript data is the standard way these methods get combined in human disease work 7.
What arrives on disk
Whole-genome sequencing gives you FASTQ or, better, aligned CRAM against GRCh38 with alternate contigs handled explicitly. A 30x CRAM is roughly 15 to 25 GB. Ask for the gVCF alongside the VCF, because a plain VCF cannot distinguish a homozygous-reference site from a site with no coverage. Annotate with Ensembl VEP, adding --everything plus a gnomAD frequency plugin. Filter to variants below 0.1 percent allele frequency before reading anything by hand, or you will drown.
RNA sequencing from PAXgene-stabilized whole blood should be globin-depleted, since hemoglobin transcripts can otherwise consume a large share of your reads. Thirty to fifty million paired-end reads is a reasonable target. We would run salmon quant with --validateMappings --seqBias --gcBias against a decoy-aware index, then import with tximport for gene-level counts. Keep the transcript-level output too, since isoform switches are often the interesting signal and are invisible after collapsing to genes.
Proteomics arrives in one of two shapes. Affinity panels such as Olink deliver NPX values, which are log2 relative units on a per-assay scale, meaning you can compare a protein to itself over time but not to a different protein. Mass spectrometry delivers mzML or vendor raw files, and the practical path is DIA-NN in library-free mode with match-between-runs enabled, producing a report.tsv you filter at one percent protein-level FDR. Expect a few hundred proteins from undepleted plasma and low thousands after depletion or a similar enrichment step.
Working with n=1
Every method above was designed for cohorts. With one person you have no between-subject contrast, so you must build the contrast from time. This is the single most important design decision, and it is why repeat sampling matters more than assay breadth.
Two approaches work. The first is a personal baseline: collect four or more timepoints under matched conditions, compute per-analyte mean and standard deviation, and score new draws as z-scores against yourself. The second is a population reference, comparing your transcript levels against GTEx whole-blood distributions to flag outliers. Use both, since the first controls for your own biology and the second catches values that are stable for you and still unusual.
Standardize the pre-analytic variables or the biology will be buried under noise. Fix the draw time to within an hour, fast consistently, and use the same tube type and the same lab. Cortisol-driven transcripts and many plasma proteins swing on a diurnal cycle large enough to dominate any subtle effect. Freeze-thaw cycles degrade the proteomic signal, so aliquot on receipt.
Where the layers disagree
Disagreement between layers is the normal case, not a sign of failure, and reading it correctly is most of the skill.
A variant with a strong annotation and no transcript change usually means the gene is not expressed in blood. Tissue specificity is the most common explanation for a silent genome finding. It is why the same pipeline gives very different answers in brain, liver, and blood 1. A transcript change with no protein change often means translational or degradation control, which is a real biological result rather than a technical one.
A protein change with no transcript change is the most interesting pattern and also the most likely to be artifact. Check hemolysis, platelet contamination, and batch first. Several proteomic platforms report per-sample quality flags, and those should gate the analysis rather than be inspected afterward.
Combining the layers is what makes any of them interpretable, and that combination is the practical definition of postgenomic analysis 8. None of this output is diagnostic. If a result points toward a clinical question, take the raw data and the report to a physician or a genetic counselor, who can order a validated clinical-grade test.
Questions people also ask
Does proteomics fall under genomics? No. Genomics measures DNA sequence, proteomics measures protein abundance and modification, and the two are separate measurement domains. They are both used within functional genomics, which is the discipline that asks what the genome does 5.
What is the difference between genomics and proteomics? Genomics is stable, digital, and amplifiable, so one blood draw is enough for life. The proteome changes hour to hour, spans about ten orders of magnitude in abundance, and cannot be amplified. Proteomics therefore needs repeated sampling and careful pre-analytic control 6.
What is functional genomics? It is the systematic study of what genes and their products do, using genome-scale readouts rather than one gene at a time. In practice that means transcriptomics and proteomics run across the whole complement of genes, then interpreted against sequence and annotation 1 3.
What are the three main types of proteomics? Expression proteomics for abundance, structural proteomics for localization and interactions, and functional proteomics for activity and post-translational modification. Blood-based personal profiling is almost entirely expression proteomics.
Footnotes
-
K E Wilson. Functional genomics and proteomics: application in neurosciences. Journal of Neurology, Neurosurgery & Psychiatry, 2004. https://doi.org/10.1136/jnnp.2003.026260 ↩ ↩2 ↩3
-
Francis S. Collins. Implications of the Human Genome Project for Medical Science. JAMA, 2001. https://doi.org/10.1001/jama.285.5.540 ↩
-
Lincoln Stein. Genome annotation: from sequence to biology. Nature Reviews Genetics, 2001. https://doi.org/10.1038/35080529 ↩ ↩2
-
Mouse Genome Sequencing Consortium. Initial sequencing and comparative analysis of the mouse genome. Nature, 2002. https://doi.org/10.1038/nature01262 ↩
-
Bent Honoré, Morten Østergaard, Henrik Vorum. Functional genomics studied by proteomics. BioEssays, 2004. https://doi.org/10.1002/bies.20075 ↩ ↩2
-
Patricia Kahn. From Genome to Proteome: Looking at a Cell’s Proteins. Science, 1995. https://doi.org/10.1126/science.270.5235.369 ↩ ↩2
-
Shuvomoy Banerjee, Juni Banerjee, Anand Krishna Tiwari. Functional Genomics and Human Diseases. Advances in Genomics, 2024. https://doi.org/10.1007/978-981-97-3169-5_7 ↩
-
S. A. Solodskikh, M. V. Gryaznova, I. D. Dvoretskaya, et al. Postgenomic technologies for genomic and proteomic analysis in biological and medical research. Ukrainian Journal of Ecology, 2019. https://doi.org/10.15421/2020_1 ↩
Pothos builds longitudinal molecular profiles of individuals: whole-genome sequencing, single-cell RNA sequencing, and plasma proteomics, delivered as a detailed report, every data file, and stored cells. Build your profile.