Documentation

Omni-MIDD Model Incubator

Omni-MIDD is a browser-based platform for mechanistic pharmacokinetic modelling across drug modalities. Models are implemented in vanilla JavaScript and run entirely client-side — no installation or server connection is required.

The platform currently supports PBPK, QSP, and PK/PD frameworks for small molecules, monoclonal antibodies, antibody–drug conjugates, PROTACs, and bispecific antibodies.

Quick Start Run your first simulation in under two minutes. models.yaml Register a new model in the library. Parameters Full reference for all model parameters. Contribute Add your own model to the incubator.

Quick Start

1. Serve the files locally

Browsers block fetch() on file:// URLs. Start a local static server from the pk_model/ directory:

# Python 3
python3 -m http.server 8000

# Node.js (npx)
npx serve .

Then open http://localhost:8000/landing.html in your browser.

2. Open a model

From the landing page, locate the model card you want to run and click Open Model. Cards marked In Development or Planned are not yet executable.

3. Adjust parameters and run

Use the slider panel at the bottom of the simulation interface to adjust model parameters. Click Run Simulation to update all charts. Use Shift + drag on any chart to zoom, and plain drag to pan.

Parameter values are preserved in localStorage between sessions. Click Reset in the parameter panel header to restore defaults.

PBPK Framework

All models in Omni-MIDD are built on a common physiologically based pharmacokinetic scaffold. Mass-balance ordinary differential equations describe drug disposition across tissue compartments connected by blood flow, with compartment volumes and blood flow rates sourced from species-specific physiological databases.

The general mass-balance equation for tissue compartment i is:

dA_i/dt = Q_i · (C_art - C_i / Kp_i) - CL_i · C_i

where Q_i is tissue blood flow, C_art is arterial drug concentration, Kp_i is the tissue:plasma partition coefficient, and CL_i is tissue intrinsic clearance.

models.yaml Reference

The model registry is defined in models.yaml. Each entry corresponds to one card on the landing page. The file is parsed at runtime by landing.js using js-yaml — no build step is required.

Field	Type	Required	Description
id	string	Yes	Unique slug used as the HTML anchor id.
title	string	Yes	Display title shown on the model card.
modality	enum	Yes	`small-molecule` \| `mab` \| `adc` \| `protac` \| `bispecific`
model_types	list	Yes	Any combination of `PBPK`, `QSP`, `PK/PD`.
status	enum	Yes	`planned` \| `in-development` \| `mature`
version	string\|null	No	Semantic version string, e.g. `1.2.0`. Null if unreleased.
drug	string	Yes	`Generic` or a specific drug name.
url	string\|null	No	Relative path to the executable model page.
description	string	Yes	One- to three-sentence summary shown on the card.
publication	object\|null	No	Bibliographic record. See sub-fields below.
publication.title	string	—	Article title.
publication.authors	list	—	Author strings in Last FM format.
publication.journal	string	—	Full journal or conference name.
publication.year	integer	—	Publication year.
publication.doi	string	—	DOI without the `https://doi.org/` prefix.

Adding a Model

To register a new model in the library, append an entry to models.yaml following the field reference above. The landing page will pick up the new card automatically on next load — no changes to HTML or JavaScript are required.

Set status: planned and url: null for models that are not yet executable. The Open button will be rendered as disabled until a url is provided.

To implement the model itself, create a new HTML page (e.g. my-model.html) that follows the same structure as index.html, sharing styles.css and referencing model-specific charts.js and params.js files.