The layers
A map of the -omes
"-omics" means measuring all, or nearly all, members of a class of molecules at once, without choosing them in advance. That shift from hypothesis-first to measure-everything changes the statistics: you test thousands of features with few samples, so multiple testing, normalisation and study design dominate.
The full list
| Omics | What is measured | Main technologies | Typical features per sample | Core question |
|---|---|---|---|---|
| DNA level | ||||
| Genomics | DNA sequence and variation | Short-read WGS/WES (Illumina, Element, Ultima, MGI), long reads (PacBio HiFi, Oxford Nanopore), SNP arrays | ~5M SNVs/indels, ~25k SVs | Which variants does this genome carry, and which matter? |
| Metagenomics | All DNA in an environmental or host sample | 16S/ITS amplicons, shotgun sequencing, long reads | 102–104 taxa | Who is there and what can they do? |
| Pan-genomics | Sequence diversity across many genomes of a species | Many de novo assemblies, graph genomes | graph nodes/edges | What does one reference miss? |
| Epigenomics | DNA methylation, histone marks, accessibility, TF binding, 3D structure | WGBS, EM-seq, arrays, nanopore 5mC; ChIP-seq, CUT&Tag, ATAC-seq, DNase-seq; Hi-C, Micro-C | ~28M CpGs; 104–105 peaks | Which parts of the genome are readable here, and why? |
| RNA level | ||||
| Transcriptomics | RNA abundance and structure (isoforms) | Bulk RNA-seq, long-read RNA-seq, microarrays (legacy), qPCR panels | ~20k genes, ~100k+ transcripts | Which genes are used, how much, in which isoform? |
| Single-cell transcriptomics | RNA per individual cell or nucleus | 10x Chromium, combinatorial indexing (Parse, Scale), Smart-seq3, BD Rhapsody | 103–106 cells × 20k genes | Which cell types and states exist, and how do they change? |
| Spatial transcriptomics | RNA with tissue coordinates | Visium/Visium HD, Stereo-seq, Slide-seq; imaging: Xenium, MERFISH/MERSCOPE, CosMx, seqFISH | 104–107 spots or cells | Where are cell states, and who are their neighbours? |
| Epitranscriptomics | RNA modifications (m6A, Ψ, m5C, A-to-I) | MeRIP-seq, miCLIP, GLORI, direct RNA nanopore | 104 sites | How are RNA fate and translation tuned chemically? |
| Translatomics | Ribosome-bound mRNA | Ribo-seq, polysome profiling, TRAP | ~10k translated ORFs | What is actually being translated, and where do ribosomes start? |
| Interactomics (RNA) | RNA–protein and RNA–RNA contacts | CLIP-seq (eCLIP, iCLIP), PARIS, RIC | binding sites | Which proteins regulate which RNAs? |
| Protein level | ||||
| Proteomics | Protein identity and abundance | LC-MS/MS (DDA, DIA, TMT), affinity (Olink, SomaScan), antibody arrays; emerging single-molecule sequencing | 3k–12k proteins (cells); 1k–11k (plasma, affinity) | Which proteins are present, how much, in what form? |
| PTM-omics | Phosphorylation, ubiquitination, acetylation, glycosylation | Enrichment + MS (TiO2, IMAC, antibody pulldown) | 104–105 sites | Which pathways are switched on? |
| Structural proteomics | 3D structure and conformations | Cryo-EM, X-ray, XL-MS, HDX-MS, limited proteolysis; AlphaFold predictions | structures | How does shape explain function and variant effects? |
| Interactomics (protein) | Protein–protein interactions and complexes | AP-MS, BioID/TurboID proximity labelling, Y2H, co-fractionation | 104–105 interactions | Which machines do proteins form? |
| Immunomics / immunopeptidomics | TCR/BCR repertoires; HLA-presented peptides | AIRR-seq, single-cell VDJ, HLA peptide MS | 105–107 clonotypes | What does the immune system see and remember? |
| Glycomics | Glycans and glycoproteins | MS, HPLC, lectin arrays | 102–103 glycans | How do sugar chains modulate function and recognition? |
| Small-molecule level | ||||
| Metabolomics | Small molecules | LC-MS, GC-MS, NMR, CE-MS, MS imaging | 102 (targeted) to 104 features (untargeted) | What is the biochemical state right now? |
| Lipidomics | Lipid species | Shotgun or LC-MS lipidomics | 102–103 lipids | How are membranes, signalling and energy storage changing? |
| Fluxomics | Reaction rates | Stable-isotope tracing (13C, 15N) + MS/NMR, flux balance models | 101–103 fluxes | How fast is material moving through pathways? |
| Exposomics | Environmental exposures over life | High-resolution MS of blood/urine, sensors, questionnaires | 103–105 features | What has this person been exposed to? |
| Organism and clinic | ||||
| Phenomics | Systematic phenotypes | EHRs, HPO terms, wearables, high-content imaging, Cell Painting | 102–103 traits | How do we describe outcomes rigorously? |
| Radiomics / pathomics | Quantitative image features | CT, MRI, PET; whole-slide histology + deep learning | 102–103 features | Can images stand in for, or add to, molecular profiles? |
| Pharmacogenomics | Genetic determinants of drug response | Genotyping panels, WGS, star-allele calling | dozens of genes | Which drug and dose for this genotype? |
| Liquid biopsy / fragmentomics | Cell-free DNA/RNA in blood | Deep targeted sequencing, low-pass WGS, cfDNA methylation | fragments, VAFs, methylation | Can we detect and track disease from blood? |
Three properties that shape every analysis
p ≫ n
Far more features (p) than samples (n). Classical regression breaks down; you need regularisation, dimensionality reduction, empirical Bayes shrinkage and false discovery rate (FDR) control (Benjamini–Hochberg, Storey's q-value).
Measurements are relative
Sequencing returns a fixed number of reads per library, so counts are compositional: one feature rising forces others down. Normalisation (size factors, TMM, CLR) tries to recover comparable scales. Spike-ins can restore absolute scale.
Technical variation is large
Batch, operator, reagent lot, run date and sample handling can explain more variance than biology. Randomise, block and record; correct afterwards only when the design allows it.
The generic analysis arc
Every layer follows a similar arc, even though the tools differ:
- Design: question, replicates, power, randomisation, controls, metadata standards.
- Primary processing: instrument signal to reads or spectra; QC (FastQC, MultiQC; PTXQC for proteomics).
- Secondary processing: reads to features (alignment, quantification, variant/peak calling, identification).
- Feature-level QC and normalisation: filter low-quality samples and features, correct library size or loading, examine batch with PCA.
- Statistical modelling: differential abundance, association, clustering, prediction.
- Interpretation: annotation, enrichment, networks, integration with other layers and with prior knowledge.
- Validation and sharing: independent cohort, orthogonal assay, data deposition (GEO, PRIDE, MetaboLights, EGA).