The read returns which ion currents can produce it, ranked, and the one experiment that would confirm the cause. On a published human iPSC-cardiomyocyte model, from the panel you already have.
Send a standard ion-channel panel. Three currents enter the model and set the score: hERG/IKr, ICaL and INa-peak. Data for other currents, including IKs and INaL, are read and reported but do not enter the simulation. We run the panel on the Kernik-Clancy 2019 human iPSC-cardiomyocyte action-potential model and return the ranked mechanisms, the risk class, and the confirming experiment.
The currents behind the signal, ranked. The per-current changes that reproduce your observation. Where more than one mechanism fits, we rank them rather than pick one.
Fractional block per current from 1 to 4x free Cmax, and the APD90 change at each step. Block enters as a static Hill function, so washout kinetics are not modeled.
The TdP-risk class your panel maps to, built from those currents through a static Hill block, the pharmacological construction CiPA applies to its ORd model.
The confirming experiment separates the surviving mechanisms. We profile the model over its conductances at your observed APD90, name which pairs that observation leaves tied, and find the measurement that unties them.
A single-cell model is a proxy for tissue-level conduction, so the read bounds the causes and the confirming experiment proves one. It states the limits of what each current resolves. Kernik-2019 carries no distinct late-sodium current, so INaL block stays out of its view.
On the 28-drug CiPA panel the free method wins 18 of 28 against our 17 of 28: a hERG-only margin built from two published numbers beats our three-class call. The loss is one compound, and the free method misses verapamil, the case a multichannel model is built for. We publish every row, including the three where the margin gets the class right and we get it wrong, on the benchmark page.
A spreadsheet gives you a class call, and on the public benchmark a free hERG margin scores slightly better than we do. What a read adds to the margin is the mechanism, the dose curve, and the confirming experiment, and you can test the format on a free read before you pay. We run the same cell type your confirmatory assay uses, so the prediction is checkable in your own lab.
Most of the value is in the first two steps. The wet-pilot design is optional; the read stands without it. No wet pilot has run for a customer yet.
The free first read takes one compound from published or non-confidential channel data. It returns the same style of report you would later pay for. Confirming-experiment design on a proprietary compound is part of the paid read.
The first read is free. A paid single-compound mechanism read starts at $12,000 and takes one to two weeks. Multi-compound packages and retainers are available, and the wet-pilot design is quoted separately.
You are not buying a risk label. You are buying the mechanism answer and the experiment that confirms it.
Knock down a cardiac gene and watch the action potential change in mV and ms, computed by a published, validated electrophysiology model (Kernik-Clancy 2019). Each channel case reproduces a known channel-disease phenotype.
Dofetilide, verapamil, ranolazine and astemizole, run against the 28-drug CiPA reference. Two land on the right class. Two do not, and the wrong ones are why the page is worth reading: verapamil because only a two-current search finds its calcium story, ranolazine because the missing current is a structural limit. One beats a free hERG margin. Each carries a confirming experiment and the limits of the read.
An annotated sample report on a public compound, section by section, with a note on each section explaining why it is there and how it fails. Same format you would get for a compound of your own.
Kernik-Clancy 2019 against the adult ventricular models, what each one is good for, and the scope of ours. Includes the reproduction check against the published baseline.
The hERG margin wins on the CiPA panel. We publish every per-drug score, both confusion matrices, and the row-by-row comparison behind the loss.
Daniel Reda. Two prior exits in life science data: CureTogether (acquired by 23andMe) and Redasoft (acquired by Hitachi). Background in Molecular Genetics.