{"entity":{"id":"multi-omic-integration","kind":"term","name":"Multi-omic integration","aka":["multi-omic integration","multi-omics integration","multiomic integration","integrative multi-omics","Multi-Omics Factor Analysis"],"tldr":"Multi-omic integration analyses measurements from different molecular layers, such as DNA, RNA and protein, side by side while keeping track of what each one measured and which specimen it came from.","summary":"A DNA change, an RNA count and a protein or phosphorylation measurement answer different questions, so integration starts with an evidence record for each measurement: the patient or experimental unit, specimen, collection time, tissue region, method, units and quality limits. Measurements from one specimen are paired; measurements from separate cohorts are not, and even adjacent tissue sections do not hold identical cells. Three reports that copy one DNA call add no new layer, and a value predicted by a model, such as RNA imputed from an image, must stay labelled as imputed beside measured values. Multi-Omics Factor Analysis (Argelaguet and colleagues, 2018) is one research method: it infers hidden factors that capture biological and technical sources of variability, separates axes shared across data types from axes specific to one, and can impute missing data; the authors applied it to 200 samples of chronic lymphocytic leukaemia profiled for somatic mutations, RNA expression, DNA methylation and ex vivo drug response. A factor is a statistical pattern, not a mechanism or a validated test. Technical variation needs its own handling: Risso and colleagues (2014) showed that usual RNA sequencing normalisation mostly accounts for sequencing depth and fails to correct library preparation and other unwanted technical effects, and where treatment time and processing method always change together, those data alone cannot separate their effects. In a proteogenomic study of 105 genomically annotated breast cancers, 77 of which gave high-quality data, pathway analysis of the phosphoproteome identified a G-protein-coupled receptor cluster that was not readily identified at the mRNA level (Mertins and colleagues, 2016), an example of layers adding different information. Correlation between layers is not a tested dependency, and a controlled intervention addresses a different question. Layers need not agree numerically, and disagreement is not automatically an error: bulk RNA from a mixed sample can be high for a gene whose protein, on a tissue image, sits mainly in normal vessel cells, so averaging the two into one confidence score would hide whether malignant cells carry an accessible target. Combining layers can reveal agreement, competing explanations, missing evidence or a sample mismatch, and it does not by itself show that a matching treatment will help: the ESMO Scale for Clinical Actionability of molecular Targets (2018) grades alterations in six levels of clinical evidence by their implications for patient management, from targets ready for routine decisions to investigational targets that need more data (Mateo and colleagues, 2018; a second version appeared in 2026). For a prediction model, a patient who appears in several layers must stay on one side of a patient-level split, and many cells from one biopsy are not many patients.","asOf":"2026-11-13","links":[{"label":"Argelaguet et al., Multi-Omics Factor Analysis, a framework for unsupervised integration of multi-omics data sets (Molecular Systems Biology 2018)","url":"https://doi.org/10.15252/msb.20178124"},{"label":"Risso et al., Normalization of RNA-seq data using factor analysis of control genes or samples (Nature Biotechnology 2014)","url":"https://doi.org/10.1038/nbt.2931"},{"label":"Mertins et al., Proteogenomics connects somatic mutations to signalling in breast cancer (Nature 2016)","url":"https://doi.org/10.1038/nature18003"},{"label":"Mateo et al., A framework to rank genomic alterations as targets for cancer precision medicine: the ESMO Scale for Clinical Actionability of molecular Targets (ESCAT) (Annals of Oncology 2018)","url":"https://doi.org/10.1093/annonc/mdy263"},{"label":"El Helali et al., The ESMO Scale for Clinical Actionability of Molecular Targets, version 2 (ESCAT v2): an evidence-based framework for ranking alterations as actionability and resistance biomarkers in precision oncology (Annals of Oncology 2026)","url":"https://doi.org/10.1016/j.annonc.2026.09.008"},{"label":"RAD51 foci as a functional biomarker of homologous recombination repair and PARP inhibitor resistance in germline BRCA-mutated breast cancer (Annals of oncology : official journal of the European Society for Medical Oncology 2018)","url":"https://doi.org/10.1093/annonc/mdy099"},{"label":"ESMO recommendations on the use of circulating tumour DNA assays for patients with cancer: a report from the ESMO Precision Medicine Working Group (Annals of oncology : official journal of the European Society for Medical Oncology 2022)","url":"https://doi.org/10.1016/j.annonc.2022.05.520"},{"label":"NCI: biomarker testing for cancer treatment","url":"https://www.cancer.gov/about-cancer/treatment/types/biomarker-testing-cancer-treatment"},{"label":"NHGRI: transcriptome fact sheet","url":"https://www.genome.gov/about-genomics/fact-sheets/Transcriptome-Fact-Sheet"},{"label":"Massively parallel digital transcriptional profiling of single cells (Nature communications 2017)","url":"https://doi.org/10.1038/ncomms14049"},{"label":"Pembrolizumab plus chemotherapy versus placebo plus chemotherapy for previously untreated locally recurrent inoperable or metastatic triple-negative breast cancer (KEYNOTE-355): a randomised, placebo-controlled, double-blind, phase 3 clinical trial (Lancet (London, England) 2020)","url":"https://doi.org/10.1016/s0140-6736(20)32531-9"},{"label":"Pembrolizumab for Early Triple-Negative Breast Cancer (The New England journal of medicine 2020)","url":"https://doi.org/10.1056/nejmoa1910549"},{"label":"Keytruda U.S. prescribing information: October 2026 effective label","url":"https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=9333c79b-d487-4538-a9f0-71b91a02b287"},{"label":"ASCO/CAP: ER and PR testing recommendations (2020)","url":"https://www.cap.org/wp-content/uploads/documents/erpgr_summaryrecommendations.pdf"},{"label":"CAP: HER2 testing guideline update (2023)","url":"https://www.cap.org/cap-guidelines/her2-testing-in-breast-cancer-2023-guideline-update/"},{"label":"The evaluation of tumor-infiltrating lymphocytes (TILs) in breast cancer: recommendations by an International TILs Working Group 2014 (Annals of oncology : official journal of the European Society for Medical Oncology 2015)","url":"https://doi.org/10.1093/annonc/mdu450"},{"label":"Comprehensive molecular portraits of human breast tumours (Nature 2012)","url":"https://doi.org/10.1038/nature11412"}],"tags":[],"related":[],"cancers":[],"sections":[],"technologies":["proteomics","spatial-transcriptomics"],"targets":[],"drugs":[],"companies":[],"institutions":[],"pathways":[],"terms":["multimodal-fusion","batch-effects","cell-composition-confound","mrna-protein-concordance","bulk-rna-seq","tumour-purity","data-leakage"],"trials":[],"people":[],"bottlenecks":[],"keyPapers":[],"journals":[],"dependsOn":[],"notes":["Integration card for combining molecular layers (a reading guide, not a score). Measures: relationships among named layers, not one universal multi-omic score. How: match evidence records, check quality and comparability, harmonise features, compare or model, then evaluate the stated question. Input and tissue cost: existing assay data and metadata; computation uses no extra tissue, but each original assay has its own specimen cost. Output and units: a comparison, factors, clusters or predictions, with the model version and underlying units. Thresholds: specific to the method and validated task, and none applies to every integration. Failure modes: sample swaps, cell-mixture changes, batch confounding, missing data and information leaking into evaluation. Cannot establish alone: causation, malignant-cell target location or better treatment outcomes. Validation tier: research integration gives neither its inputs nor its resulting model a clinical status.","Figure reading (fictional): the same DNA call, a T in the sequence A C T G A, is copied into Report A, Report B and Report C, which is one observation reported three times. A separate RNA assay showing a splice junction is a new measurement that answers a new question, and it does not by itself confirm the DNA call.","Figure reading (fictional Target Q): specimen S at time T1 supplies bulk RNA, where Q is abundant, and a tissue image, where Q protein is seen in normal vessel cells. Whether malignant cells carry accessible Q is unresolved, a model-imputed Q is labelled as an estimate rather than a measurement, and a different specimen or time point is not paired evidence.","Genome, transcriptome and proteome name DNA, RNA and protein collections. An omics profile measures many features within a layer, subject to its coverage and capture limits; it does not necessarily inspect every molecule. Molecular layer and spatial or cellular resolution answer different questions. Multi-omic integration relates the measured layers, with their specimen and analysis context retained, rather than turning a larger number of assays into a universal measure of certainty.","Consider an invented repair-gene R example, with measurements attributed to the same pretreatment tissue region except for a derived model. A supported splice-site DNA variant identifies a candidate sequence change. Loss of the other allelic contribution, with the variant assigned to the retained contribution, supports a possible two-hit explanation subject to copy and phase uncertainty. A reproducible abnormal RNA junction in malignant-cell-enriched material adds evidence for a changed transcript product, while reduced staining with suitable controls adds a protein-signal difference in the assigned cells. Neither establishes that every intact protein is gone. A high genomic scar adds evidence about genomic history, not a direct test of present repair. Impaired repair in a suitable model with informative cycling and damage controls addresses function under those conditions. A replicated drug response at specified exposures adds a model-sensitivity hypothesis; it still leaves achievable safe exposure and patient benefit to their appropriate evidence. These are distinct observations, not a requirement to perform every assay before using any established clinical biomarker.","A practical report reading starts with six questions: what specimen, region, date and treatment context were sampled; what object and units were measured with which assay version; whether the result is signed clinical, preliminary, research or model-predicted evidence; which controls or reference support the comparison; what remains unmeasured; and which actual decision could change. A missing field is a question to resolve, not a fact to assume. For fictional pretreatment bulk RNA, high R relative to a stated reference supports captured expression in that mixture. Calling it a surface target remains an interpretation until the contributing cells and accessible protein are established; selectivity, dependence and treatment benefit need their own evidence. Molecular layer and resolution are separate axes. Mapping 500 RNA targets in one section adds location relative to a pooled RNA result, while the measured layer remains RNA. It does not add a membrane-protein or functional measurement. In that invented example, one pretreatment breast-biopsy region contains malignant cells and normal neighbours; no other disease site was sampled. R is reported in the assay's normalised units against a compatible reference, so high belongs to that comparison. Another broad RNA profile need not resolve the missing cell or membrane-protein location. The result can focus research follow-up; treatment selection still needs the validated biomarker-treatment link, clinical setting and eligibility.","An entirely fictional core-biopsy report lists invasive carcinoma of no special type, grade 3; ER and PR below 1 percent; HER2 IHC one-plus; Ki-67 positivity of 60 percent; stromal TILs of 30 percent; PD-L1 22C3 combined positive score 15; and a supported TP53 missense finding at VAF 38 percent on a tumour-only panel. These observations retain separate rules and denominators. The diagnosis describes the sampled region; the receptor results support a triple-negative category and low HER2 staining under the stated testing rules, without establishing a particular drug benefit. Ki-67 positivity is not a treatment-selection rule here. The stromal score is an eligible-area assessment, not proof of recognition, and CPS is its own defined stain score rather than a universal immune percentage. VAF is not inherited status, cancer-cell fraction or measured variant function. KEYNOTE-355 tested pembrolizumab plus chemotherapy in previously untreated advanced TNBC; its CPS-at-least-ten group had median PFS 9.7 versus 5.6 months, hazard ratio 0.65. KEYNOTE-522 enrolled high-risk early TNBC regardless of PD-L1 status. A score cannot transfer benefit between those settings. None of these invented values is a patient observation or a prescription."],"provenance":{"editedBy":"OnCo systematic education audit, round 3 (spatial and computational content): Crossref and Europe PMC identity checks, abstracts and selected full texts read","editedOn":"2026-11-13","note":"Argelaguet 2018 (PMID 29925568), Risso 2014 (PMID 25150836) and Mertins 2016 (PMID 27251275): DOI and PMID resolved and matched on title, authors and year; the 200-sample leukaemia application, the biological-and-technical factor description, the depth-only normalisation finding and the 105/77 breast cancer counts were read in the abstracts. The paired-versus-unpaired, imputation-label and patient-level-split points are the lesson's methodological principles, not findings of those papers. The NCI biomarker-testing page was re-read in selected complete paragraphs on 13 November 2026: its matching-treatment discussion says that a matching biomarker does not ensure benefit. This supports the qualified evidence boundary here; no complete current test inventory or whole-page clinical verification is claimed. The report-card example is fictional."},"category":"Methods and models"},"route":"/terms/multi-omic-integration/","neighbours":{"technology":[{"id":"proteomics","kind":"technology","name":"Proteomics & phosphoproteomics","route":"/technologies/proteomics/"},{"id":"spatial-transcriptomics","kind":"technology","name":"Spatial transcriptomics","route":"/technologies/spatial-transcriptomics/"}],"term":[{"id":"batch-effects","kind":"term","name":"Batch effects and harmonisation","route":"/terms/batch-effects/"},{"id":"bulk-rna-seq","kind":"term","name":"Bulk RNA sequencing (RNA-seq) and the full transcriptome","route":"/terms/bulk-rna-seq/"},{"id":"cell-composition-confound","kind":"term","name":"Cell composition confound (tumour versus stroma and immune cells)","route":"/terms/cell-composition-confound/"},{"id":"data-leakage","kind":"term","name":"Data leakage in model evaluation","route":"/terms/data-leakage/"},{"id":"mass-spec-proteome","kind":"term","name":"Deep mass-spectrometry proteome (CPTAC)","route":"/terms/mass-spec-proteome/"},{"id":"mrna-protein-concordance","kind":"term","name":"mRNA to protein concordance","route":"/terms/mrna-protein-concordance/"},{"id":"multimodal-fusion","kind":"term","name":"Multimodal fusion (early, late, modality dropout)","route":"/terms/multimodal-fusion/"},{"id":"tumour-purity","kind":"term","name":"Tumour purity","route":"/terms/tumour-purity/"}]}}