Skip to content

The zebrafish VAMR behaviour assay and where it meets the pathograph

The Visual and Acoustic Motor Response (VAMR) assay measures how zebrafish larvae move in response to light changes and tones. Spath et al. (2026) ran it on 17 reference chemicals for developmental neurotoxicity (DNT) and compared the results with the cell-based DNT in vitro battery (DNT IVB). This report records the assay's design, the per-chemical results, and the dismech nodes each tested chemical could attach to.

No dismech entry records this assay. Adding it is tracked in issue #13450.

Read on 2026-10-03; kb/ as of commit 093f725c24.

Sources

Source Identifier What it gives
Article doi:10.1016/j.neuro.2026.103414, PMID:41780647, NeuroToxicology 114:103414, CC BY PubMed holds the abstract only; not in PMC
Open full text url:https://zenodo.org/records/23035979/files/1-s2.0-S0161813X26000355-main.pdf just fetch-reference on this URL caches the whole article, so it is the citable form for anything beyond the abstract. The cache file is committed by #13451, not by this PR
Supplementary workbook https://ars.els-cdn.com/content/image/1-s2.0-S0161813X26000355-mmc3.xlsx Tables S1–S7. S3 and S4 hold every concentration-response fit
Dataset DOI printed in the article 10.5281/zenodo.17937513 Not registered; does not resolve

data/vamr-endpoint-hit-calls-2026-10-03.tsv is Tables S3 and S4 reduced to one row per chemical, exposure arm and endpoint (884 rows), with a column saying whether the row passes the article's hit rule.

Regenerate it, or check it against the workbook, with scripts/vamr_endpoint_hit_calls.py:

uv run python scripts/vamr_endpoint_hit_calls.py --check   # downloads the workbook
uv run python scripts/vamr_endpoint_hit_calls.py --out docs/reports/data/vamr-endpoint-hit-calls-2026-10-03.tsv

The workbook itself is not committed: it is a publisher asset, and while the article is CC BY the supplement's own terms are not stated. The script downloads it, so the numbers below can be re-derived rather than trusted.

Assay design
Organism Zebrafish (Danio rerio), in-house UFZ-OBI/WIK strain
Format One larva per well, 96-square-well polystyrene plate, 400 µL 10% Hanks' balanced salt solution, 28 °C, 14:10 h light:dark
Acute arm Larvae plated at 4 days post-fertilization (dpf), exposed at 5 dpf for 40 minutes, then tested. 36 larvae per concentration, 72 controls. Six concentrations, quarter-log spacing, 4.4–80 µM, extended downward for potent chemicals
Developmental arm Embryos plated and exposed once at 1 dpf, tested at 5 dpf. 32 larvae per concentration, 96 controls. Nine concentrations, semi-log spacing, with the top concentration set below the level causing malformation
Vehicle 0.4% DMSO
Stimuli Light at 13.5 klux against dark at 0 lux; 300 Hz tones at 65 dB (low) and 75 dB (high)
Instrument ViewPoint Zebrabox, ZebraLab software in quantization mode, 25 frames per second
Readout Pixel change per second
Run time 1 h 17 min
Exclusions Malformed larvae and larvae without an inflated swim bladder are removed before analysis

The 26 endpoints

Group Endpoints What is measured
Visual startle VSRB, VSR1, VSR2 Movement in the 1 s after a light change
Baseline locomotion BSL1–4 Four 5-minute blocks in the dark
Visual motor response VMR1; VMR2–5 10 minutes in the light; four 5-minute blocks back in the dark
Acoustic startle ASR1, ASR2, ASR3 Mean response to five low tones, five high tones, and five high tones after habituation training
Habituation learning ASH1 Decline in response within the first bout of 30 high tones at 1 s intervals
Potentiation of habituation ASH1/5 Fifth bout relative to the first and fifth together
Habituation total ASHsum Summed activity over the five bouts
Memory retention ASR2/3 ASR2 relative to ASR2 and ASR3 together
Rest-period activity ISI1–3, IEI1–3, IBI Movement between stimuli, between endpoints, and between habituation bouts

Analysis and hit rule

  • Time-series endpoints (BSL1–4, VMR1–5) are fitted with generalized additive mixed models. The Fréchet distance between each exposed group's fitted curve and the control curve is the response.
  • Single-value endpoints are fitted with Gaussian linear models. The absolute difference from the control mean is the response.
  • Both feed the US EPA tcplfit2 pipeline, which returns AC50, a point of departure and a benchmark concentration.
  • An endpoint is a hit when its hit call is at least 0.9 and its curve fit carries fewer than two flags.

Because both response measures are absolute, the tables carry no direction. Whether a chemical raised or lowered movement is stated only in the article's text and figures.

Results

Twelve chemicals with human or rodent evidence of DNT and five predicted negatives were tested. Three of the twelve (phenytoin, maneb, nicotine) are negative in the DNT IVB.

Acute Developmental
DNT-positive chemicals detected, as stated in the text 11 of 12 8 of 12
The three DNT IVB false negatives all detected none detected
Predicted negatives called negative, as stated in the text 5 of 5 4 of 5 (ibuprofen is a hit)
Potency against the DNT IVB no significant difference 1–4 orders of magnitude lower AC50

Both rows above are the article's own figures. Applying the hit rule to the supplementary tables gives 9 positives with at least one passing developmental endpoint rather than 8; the extra one is ketamine, and the discrepancy is set out below.

The authors state that acute hits "likely reflect short-term, receptor-mediated effects" and that it is not clear whether they predict developmental outcomes.

Hits per chemical

Counts are endpoints passing the hit rule, out of 26, taken from the supplementary tables. "Most sensitive" is the passing endpoint with the lowest AC50, which is how the paper's own most-sensitive column can be checked against these. AC50 in µM.

The two hit columns are not produced by the same rule, so do not read a difference between them as a difference in effect. The rule has two halves — a hit call of at least 0.9, and fewer than two curve-fit flags — and only the first half does any work in the developmental arm. There, 71 rows clear the hit call and none of them carries a single flag, so the flag half excludes nothing. In the acute arm, 121 rows clear the hit call and the flag half excludes 11 of them. The developmental column is therefore effectively hitcall >= 0.9 alone. Table S4 records no flags on any row above the threshold, so whether that reflects better-behaved fits or flags that were not published cannot be told from the workbook.

Chemical Class in the study Acute hits Most sensitive (acute) Developmental hits Most sensitive (developmental)
Chlorpyrifos positive 17 BSL2, 0.53 14 VMR5, 0.0010
Trichlorfon positive 1 BSL4, 3.1 15 VMR3, 0.0023
Triethyltin bromide positive 22 VMR1, 0.13 11 ASHsum, 0.0024
Haloperidol positive 22 ASHsum, 0.52 8 BSL3, 0.022
Tebuconazole positive 17 BSL4, 6.3 4 ASHsum, 0.19
Hexachlorophene positive 7 ASHsum, 0.060 2 BSL1, 0.00026
PFOA positive 1 VMR1, 11.9 1 VMR5, 0.011
Ketamine hydrochloride positive 10 BSL4, 15 2 (see below) BSL3, 40
BDE-99 positive 0 – 3 BSL4, 0.0065
Phenytoin (5,5-diphenylhydantoin) positive, DNT IVB negative 7 BSL4, 0.084 0 –
Maneb positive, DNT IVB negative 4 ASR3, 0.76 0 –
Nicotine positive, DNT IVB negative 2 VMR2, 5.6 0 –
Ibuprofen predicted negative 0 – 3 VMR2, 0.15
Saccharin predicted negative 0 – 6 (see below) BSL3, 0.0013
D-Mannitol predicted negative 0 – 1 (see below) BSL4, 0.00052
Sodium benzoate predicted negative 0 – 1 (see below) VMR5, 0.099
Acetaminophen predicted negative 0 – 0 –

Learning and memory endpoints that pass the hit rule

Endpoint Hits (AC50, µM)
ASH1, habituation Haloperidol, developmental, 1.25
ASH1/5, potentiation of habituation Phenytoin, acute, 18.3; haloperidol, acute, 10.4; hexachlorophene, acute, 0.13; trichlorfon, developmental, 1.65; triethyltin, developmental, 0.14
ASHsum, total activity across the five habituation bouts Hexachlorophene, acute, 0.060; haloperidol, acute, 0.52; triethyltin, acute, 2.1; chlorpyrifos, acute, 4.4; ketamine, acute, 32; tebuconazole, acute, 40; phenytoin, acute, 40; triethyltin, developmental, 0.0024; saccharin, developmental, 0.0047; trichlorfon, developmental, 0.021; chlorpyrifos, developmental, 0.10; tebuconazole, developmental, 0.19
ASR2/3, memory retention Phenytoin, acute, 16.2; haloperidol, acute, 45.8; hexachlorophene, acute, 0.30; tebuconazole, acute, 59; triethyltin, acute, 4.8; ibuprofen, developmental, 4.5

ASHsum has as many passing fits as the other three put together — 12 against 12 — and it is the weakest of them as a learning readout: it sums motor activity across the habituation bouts, so a chemical that simply changes how much a larva moves will move it without any change in the rate of habituation. ASH1 and ASH1/5 are the ratio endpoints that isolate the decrement itself, and between them they pass on six fits.

Where the article's text and its supplement disagree

  • Developmental negatives. The text classifies saccharin, D-mannitol, sodium benzoate and acetaminophen as negative after developmental exposure. Table S4 has rows with a hit call of at least 0.9 for saccharin (6), D-mannitol (1) and sodium benzoate (1), and Table S6 lists a developmental AC50 for each. The flag column of Table S4 is empty for every row with a hit call of at least 0.9, so the flag half of the hit rule cannot be applied to the developmental arm from the published table.
  • Ketamine, developmental. The text says ketamine was a hit only in the acute arm. Table S4 has two developmental rows at hit calls 0.907 and 0.947, both with an AC50 at 40 µM and no flags recorded.
  • Ketamine and habituation. The discussion says acute ketamine "blocked habituation learning". The acute ASH1 fit has a hit call of 0.994 and three flags, so it is not a hit under the article's rule. The acute ketamine hits are BSL1–4, VMR2–5, ASHsum and IBI, between 15 and 43 µM.
  • Nicotine and habituation. The discussion says acute nicotine impaired habituation, and Table S6 gives ASH1 at 0.22 µM as nicotine's most sensitive acute endpoint. That fit carries three flags. The two acute fits that pass are VMR2 (5.6 µM) and VSR2 (5.7 µM).
  • Trichlorfon and habituation. The discussion says acute trichlorfon impaired habituation at the highest concentration. The acute ASH1 hit call is 0.008.
  • Count of positives detected in both arms. One results paragraph says all nine DNT IVB-positive chemicals were hits in both arms; later paragraphs say seven. The tables give seven: chlorpyrifos, haloperidol, hexachlorophene, PFOA, tebuconazole, trichlorfon and triethyltin bromide.

Attachment to dismech

The assay is a whole-organism animal model, so it belongs under animal_models, observing at ORGANISM scale. It is not specific to any chemical or disease; each tie below runs through one tested chemical.

Chemicals with a node to attach to

Chemical Entry or module Node What the article supports
Chlorpyrifos Organophosphate_Poisoning Central cholinergic overstimulation Acute exposure raised movement in normally quiet periods (AC50 2.46–9.01 µM on the endpoints shown); developmental exposure lowered movement on most endpoints (AC50 0.08–0.17 µM on the same endpoints)
Trichlorfon Organophosphate_Poisoning Central cholinergic overstimulation Mostly developmental: 15 hits from 0.0023 µM
Ketamine nmda_receptor_hypofunction NMDA Receptor Antagonism or Receptor Loss Ketamine is the perturbation; the hits are locomotor. A readout on Learning and Memory Impairment is not supported by a passing fit
Nicotine Organophosphate_Poisoning; Green_Tobacco_Sickness Nicotinic receptor overstimulation; Nicotinic Acetylcholine Receptor Overstimulation Two acute hits on responses to light; none developmental

The article measures behaviour only. It does not measure acetylcholinesterase activity, receptor occupancy or any molecular quantity, and it does not attribute the organophosphate results to cholinesterase inhibition. Lead_Poisoning conforms to three nodes of the NMDA module, but lead was not tested.

Diseases the article names without data

The introduction cites attention deficit hyperactivity disorder and autism spectrum disorder as outcomes of developmental neurotoxicity. Attention_Deficit-Hyperactivity_Disorder has no environmental entries. The article cites Dalsager et al. 2019 (doi:10.1016/j.envres.2019.108533) for maternal chlorpyrifos and pyrethroid metabolites and ADHD symptoms in young children.

Matching readouts with no tested chemical behind them

  • Exaggerated Startle Response is a phenotype in ten entries, among them Hereditary_Hyperekplexia, Stiff_Person_Syndrome, Tay-Sachs_Disease and Sandhoff_Disease. ASR1–3 measure startle magnitude, but none of the 17 chemicals bears on those diseases.
  • The introduction describes the assay as able to detect seizure-like activity. This article reports no seizure-like result, so it gives no tie to epilepsy_excitation_inhibition_imbalance.

What dismech lacks

  • No pathophysiology or phenotype node is named for habituation.
  • No entry or module mentions maneb, PFOA, hexachlorophene, triethyltin, tebuconazole or trichlorfon.
  • Chlorpyrifos is named once, in the description of Organophosphate_Poisoning. That entry has no animal, experimental or computational model.
  • No entry is named for fetal hydantoin syndrome.