DNA, RNA, proteins and metabolites offer different perspectives on biology. Multi-omics integrates information from at least two omics disciplines, helping researchers investigate how those measurements relate to one another and to a defined research question.
The purpose goes beyond producing separate lists of genetic variants, RNA levels or proteins. Analysing relationships across datasets can reveal patterns that would be harder to interpret from a single molecular layer.
Which biological layers can be combined?
Genomics examines DNA, including genetic variation. Transcriptomics studies RNA, providing information about gene expression in a particular biological context. Proteomics investigates proteins, while metabolomics measures small molecules involved in metabolism, such as sugars, amino acids and lipids. Epigenomics adds information about features such as DNA methylation that can influence gene regulation.
RNA abundance, for example, is distinct from the amount of the corresponding protein present and from that protein’s activity.
Multi-omics also extends beyond sequencing: protein and metabolite measurements often use other technologies, including mass spectrometry.
What makes measurements comparable?
Decisions about Sample Processing belong alongside assay selection, so researchers can plan the specimen portions and preparation methods each measurement requires. Separate portions, known as aliquots, may be needed for different assays. Their records should preserve the connection to the original specimen and its biological context.
Two measurements can belong to the same participant yet come from different tissues or visits. Blood collected before treatment and tissue collected afterwards could help investigate a treatment response, but they describe different biological states.
Records need to distinguish participant, tissue, collection time and specimen identity, alongside relevant treatment histories and participant characteristics. The required alignment depends on the question.
Shared specimens are one option; researchers can also integrate datasets from different people or experiments. Such designs need analytical approaches that account for unpaired observations and differences in biological context.
What integration adds: a simple example
Consider a hypothetical study comparing two groups of tissue samples using transcriptomics and proteomics. The RNA analysis identifies several genes with higher transcript levels in one group. Are the corresponding proteins also more abundant?
Researchers examine both datasets to see whether the genes and proteins showing differences point towards the same molecular pathways. Where measurements come from paired samples, they can also assess whether RNA and protein levels vary together across individual samples. Agreement would support a hypothesis that the observed difference extends across both layers.
Disagreement gives researchers a different lead. If RNA levels differ while protein levels remain similar, that discrepancy gives them a reason to investigate regulation, measurement timing or assay sensitivity.
Even agreement needs explanation. A difference in the mix of cell types within the tissue could contribute to both patterns.
Three checks before interpreting the results
Choose layers that serve the research question. Adding metabolomics to an RNA study should have a purpose, such as investigating whether changes in gene expression accompany differences in metabolic pathways. Every additional dataset brings analytical work and opportunities for chance findings. A carefully chosen pair can be more informative than a larger collection assembled without a clear rationale.
Check which samples are represented across layers. If measurements are missing, establish which samples lack which layer and why. Restricting an analysis to participants with complete data can change the composition of the study population, affecting how its findings should be interpreted.
Examine technical explanations for the pattern. Batch effects are systematic differences associated with measurement runs, reagent lots or other technical conditions. Measuring all samples from one study group in one batch and the comparison group in another makes biological and technical differences difficult to separate. Distributing study groups across batches and recording relevant metadata helps researchers assess this problem. Normalisation and statistical adjustment can address technical variation, but their effectiveness depends on the design and available data.
Cross-layer associations alone do not establish causation; demonstrating a mechanism may require follow-up experiments.
A useful multi-omics analysis brings complementary measurements into a coherent explanation, investigates why results agree or differ, and helps identify the next experiment worth doing.
