{"entity":{"id":"controlled-perturbation","kind":"term","name":"Controlled perturbation experiment","aka":["perturbation experiment","vehicle control","positive control","negative control","rescue experiment","mechanism control"],"tldr":"Controlled perturbations deliberately change specified factors and compare responses with suitable controls. A model response alone does not prove its mechanism or patient benefit.","summary":"A controlled perturbation deliberately changes a defined factor and measures the response against appropriate comparison conditions. Vehicle, positive, negative and mechanism controls answer different questions. A reproducible response can support an experimental claim without establishing its proposed mechanism or patient benefit. Baseline growth, timing, dose and the actual unit assigned to treatment matter.","asOf":"2026-11-14","links":[{"label":"Hafner et al.: growth rate inhibition metrics correct for confounders in measuring drug sensitivity (Nature Methods 2016)","url":"https://doi.org/10.1038/nmeth.3853"},{"label":"Measuring cancer drug sensitivity and resistance in cultured cells (Current Protocols in Chemical Biology 2017)","url":"https://doi.org/10.1002/cpch.21"},{"label":"The promise and peril of chemical probes","url":"https://doi.org/10.1038/nchembio.1867"},{"label":"What exactly is 'N' in cell culture and animal experiments?","url":"https://doi.org/10.1371/journal.pbio.2005282"},{"label":"Searching for Drug Synergy in Complex Dose-Response Landscapes Using an Interaction Potency Model (2015)","url":"https://pubmed.ncbi.nlm.nih.gov/26949479/"},{"label":"SynergyFinder: a web application for analyzing drug combination dose-response matrix data (2017)","url":"https://pubmed.ncbi.nlm.nih.gov/28379339/"},{"label":"Mutant p53 disrupts mammary tissue architecture via the mevalonate pathway (2012)","url":"https://pubmed.ncbi.nlm.nih.gov/22265415/"},{"label":"Guidelines for the use and interpretation of assays for monitoring autophagy (4th edition)&lt;sup&gt;1&lt;/sup&gt; (2021)","url":"https://pubmed.ncbi.nlm.nih.gov/33634751/"},{"label":"Proof-of-concept study linking ex vivo sensitivity testing to neoadjuvant anthracycline-based chemotherapy response in breast cancer patients.","url":"https://pubmed.ncbi.nlm.nih.gov/37777518/"}],"tags":[],"related":["preclinical","in-vitro-in-vivo","cell-line","organoid","placebo"],"cancers":[],"sections":[],"technologies":[],"targets":[],"drugs":[],"companies":[],"institutions":[],"pathways":[],"terms":[],"trials":[],"people":[],"bottlenecks":[],"keyPapers":[],"journals":[],"dependsOn":[],"notes":["A vehicle control contains the carrier without the active treatment; an untreated condition asks a different question if the carrier has effects. A positive control checks that the assay can detect a known relevant response, while a negative or non-targeting control estimates background under its design. Another independent way to disrupt the same function can support a mechanism, and a rescue restores a function to test whether that reverses the response here. Off-target effects, incomplete restoration and model context limit interpretation. These are questions for matched model, timing and readout; no one control proves specificity or patient benefit, and agreement between similar readouts is not automatically independent mechanistic confirmation.","Keep what changed, the comparator, concentration or dose range, timing, assignment, readout and units, sample identities, replicate structure and variability. Repeated wells address variation under their preparation and allocation; independently obtained biological samples address another source of variation. Many wells from one tumour do not become independent patients, and repetition or a small p-value does not repair a missing comparison. A two-drug experiment needs vehicle, both single drugs and the combination. A larger combination effect than either single drug is not automatically synergy; that claim needs a stated reference model.","Fictional direct counts: vehicle-treated cells increase from 100 to 200 while drug-treated cells remain at 100. The endpoint drug count is half the vehicle count without measured net decline from the starting count, so calling it 50 per cent killing overstates the observation. Slower growth, death and a molecular change require their own readouts; direct death or recovery measurements could address further questions. Hafner and colleagues showed that conventional IC50 and maximal-effect measures depend on cell divisions during an assay and developed growth-rate metrics to address that dependence. A real experimental response does not by itself establish the responsible target or clinical benefit.","A fictional two-drug table normalises remaining viability to vehicle at 100 per cent: A leaves 70, B leaves 80, and combination examples leave 56 or 30. Highest single agent uses the better single-drug effect, a 70 per cent remaining-viability reference here. Bliss independence predicts 0.70 times 0.80 = 0.56, or 56 per cent. Both combinations exceed the HSA effect; 56 matches Bliss while 30 exceeds it at that dose pair. A further combination leaving 65 exceeds HSA but not Bliss. Loewe additivity instead needs equivalent-dose information from suitable individual dose-response curves; this one-point table cannot establish it. Report the model, dose-pair pattern, uncertainty and schedule. Interaction relative to a reference does not identify a biological mechanism, tolerable human exposure or a treatment choice.","A rescue restores a proposed missing function or supply and tests whether a defined effect reverses. Adding a cake mix back after its removal implicates the mixture without identifying flour, sugar or another ingredient; likewise, broad mevalonate add-back does not isolate cholesterol. More specific branch perturbations and appropriate add-backs can narrow the explanation. Partial rescue can reflect incomplete delivery or restoration as well as additional contributors, so it does not prove a second cause. Place the add-back relative to the blocked step: supplying GGPP after an upstream synthesis block is different from inhibiting the transferase that uses GGPP, which the added substrate cannot simply be assumed to bypass. Include restoration-only and matched controls, verify the restored process and distinguish growth, morphology and death. A genetic perturbation also has its own specificity limits.","Attribute each readout to the relevant cells. An ATP-based signal can change through cell number or per-cell metabolism, and slower single-cell mass accumulation does not by itself show death. Direct death measurements and regrowth after washout ask further questions; regrowth concerns surviving cells, not revival of cells already dead. A mixed culture can change composition, and a nonmalignant-cell label alone does not establish a validated healthy-tissue control. Ladan's breast-slice chemotherapy assay was compared with response during preoperative treatment in a defined setting. That assay-specific validation does not validate every postoperative combination, human exposure or clinical decision. An inadequate viable preparation should remain uninterpretable rather than be called resistant.","An unchanged total count can occur while some cells divide and others die. The fictional 100-to-100 treated count establishes no net change. Measuring the fraction that died requires a suitable death readout and denominator."],"category":"Methods and models"},"route":"/terms/controlled-perturbation/","neighbours":{"term":[{"id":"cell-line","kind":"term","name":"Cell line","route":"/terms/cell-line/"},{"id":"in-vitro-in-vivo","kind":"term","name":"In vitro and in vivo","route":"/terms/in-vitro-in-vivo/"},{"id":"organoid","kind":"term","name":"Organoid","route":"/terms/organoid/"},{"id":"placebo","kind":"term","name":"Placebo","route":"/terms/placebo/"},{"id":"preclinical","kind":"term","name":"Preclinical","route":"/terms/preclinical/"}]}}