A 6-Month Self-Study Course

Biological
World Models

A 6-month technical course for applying AI, mathematics, physics, and computer science to systems biology and drug discovery.

Start the Course View 6-Month Roadmap

The Central Idea

Biology is a noisy, adaptive, multiscale dynamical system.

Treat biology as a partially observed, stochastic, multiscale dynamical system. Treat drugs, CRISPR edits, knockdowns, overexpression, dose, time, and environment as interventions. The goal is to learn how to model, predict, and control biological state transitions.

Generative model of cell-state dynamics

x_t+Δt ∼ p_θ (x_t+Δt | x_t, intervention, cell type, env)

state · model · current state · intervention · context

Drug discovery becomes the problem of choosing interventions that move a biological system from a disease state toward a healthier state — while respecting toxicity, selectivity, delivery, and manufacturability constraints.

Biology as a Dynamical System

States evolve, attractors form, perturbations push trajectories. Model cells with the same vocabulary you already use for physics and ML.

Pathways as Causal Graphs

Pathway databases are priors, not executable truth. Turn signed digraphs into causal, mechanistic, learnable simulators.

Drug Discovery as Control

A drug is not just a molecule that binds. It is an intervention that changes a biological trajectory under constraints.

Featured Example · Applied

What this looks like in practice.

A real-world workbench built on the same vocabulary this course teaches: condition-first targeting, pathway-aware generation, and formal mechanism checks before anything reaches a wet lab.

Peptiter/ DiscoveryLab

CASE STUDY · 01

A peptide discovery workbench that proves what its candidates are allowed to claim.

DiscoveryLab turns therapeutic hypotheses into receptor-aware peptide families, verifies pathway mechanisms with graph checks and Lean 4 audit artifacts, and ranks candidates through claim-bounded in-silico assessment before handing testable batches to wet-lab partners.

Search reduction

10¹³ → 10²

per program

Seed library

16 parents

natural + xeno

Mechanism audit

Lean 4

package plan

State space

Sequences as discrete states; receptor binding as observable.

Intervention