CARDIE
In-silico proarrhythmia screen on the FDA's ORd-CiPA v1.0 myocyte
1.00× Cmax
Human heart
Whole-heart view
—
Generated for the current view, compound and dose — changes with everything you click.
ORd-CiPA v1.0Human ventricular myocyte model
FDA CiPA12 training compounds

Model & data sources

What do these terms mean?

Cell model & protocol

O'Hara–Rudy / ORd-CiPA is a mathematical model of a single human heart-muscle cell, built from real patch-clamp measurements. It doesn't simulate a whole heart — it solves the equations for one cell's electrical cycle, many times over.

Endo / Epi / Mid
the three layers of the ventricular wall (inner, outer, middle). Cells differ between them — see the Transmural heterogeneity panel.
Cycle length
time between simulated heartbeats. 2000 ms is a deliberately slow 30 bpm, the FDA's own CiPA protocol.
Pre-paced beats
the simulation runs 400 beats before recording, so the cell settles into its steady rhythm instead of being captured mid-"warm-up".
hERG & drug binding

hERG is a potassium channel gene; it's the single most common place a drug accidentally causes dangerous heart rhythms, because blocking it slows the heart's electrical "reset" between beats.

A Markov binding model describes how a drug molecule attaches to and detaches from hERG over time. Some drugs bind and release quickly; others get briefly "trapped" inside the channel, which is the more dangerous pattern.

Ion channels (INa, ICaL, Ito…)

Each channel is a separate "gate" in the cell membrane that lets one type of ion flow in or out. Together they draw the action-potential shape in the chart above.

INa
fast sodium — the initial spike of each beat.
INaL
late sodium — a small lingering leak; if it doesn't fade, the cell stays activated too long.
ICaL
L-type calcium — shapes the plateau and triggers the muscle to actually contract.
Ito
transient outward potassium — the brief early "notch" right after the spike.
IKr / IKs
rapid/slow delayed-rectifier potassium — the main "reset" currents; hERG encodes IKr.
IK1
inward-rectifier potassium — holds the cell at rest between beats.
qNet & ΔAPD90 (the metrics)
qNet
the net electrical charge moved across the membrane by six currents over one beat. The single number this project uses to predict risk — more negative means more risk.
APD90
how long the cell stays electrically activated before 90% resetting. A prolonged APD90 is the classic early-warning sign of risk, though a less reliable predictor than qNet (shown elsewhere on this page).
FDA CiPA
the FDA's Comprehensive in vitro Proarrhythmia Assay initiative — the regulatory framework this whole simulation implements.
descend to a single ventricular myocyte
0 ms

Action potential — dosed vs drug-free control (ghost)

Net current — shaded area is qNet

Transmural heterogeneity

Endo/Epi/Mid are three real cell-layer solves, not a tissue model. Dispersion between them — especially how much longer the M cell stays depolarised — is the actual re-entry mechanism behind torsade de pointes. Illustrative alongside the validated qNet classification, not a replacement for it.

Cross-drug comparison

Membrane potential of all 12 compounds at one moment in the beat, at the current dose. Colour is the same voltage ramp as the myocyte above; the border is the FDA risk class. Click a cell to inspect that compound.
150 ms · plateau

About

Verdict

—
qNet
—
µC/µF · control 0.0781
ΔAPD90
—
vs 277.2 ms
hERG drug-bound
—
peak occupancy
FDA label
—
CiPA training set

Channel block at this dose

Why

—
Generated from the computed block profile. The language model never touches the numerical path; narration is Gemini reading this same finished sentence aloud, not generating anything new.

Ask

Any molecule predicted

Validation — 12 CiPA training compounds, scored against FDA labels

—qNet, mean over 1–4× Cmax
—ΔAPD90 at 4× Cmax