DNA Self-Assembly
DNA Nanotechnology — Programmable Self-Assembly · Nanotechnology
Program matter to fold itself — design DNA sequences that self-assemble into a target nanostructure at high yield.
📋 The problem
DNA can be programmed to fold itself into shapes (origami) and machines. Designing sequences that self-assemble into a target structure at high yield is programmable matter in practice.
🧗 Why it's a grand challenge
Folding yield is set by a rugged thermodynamic/kinetic landscape; sequence design is combinatorial; off-target kinetic traps lower yield unpredictably.
🧮 Governing model
ΔG = Σ ΔG_NN; yield = Z_target / Z_total over the folding landscapeSequence-programmed self-assembly: hybridization thermodynamics (nearest-neighbor ΔG) set a folding free-energy landscape; correct-fold yield from the equilibrium partition function over the staple/scaffold design.
Current best: scaffolded DNA origami (caDNAno/oxDNA design pipelines)
🧭 Possible approaches
- Graph / sequence models predicting folding yield
- Generative inverse design of staple sets
- Kinetic (oxDNA-informed) trap avoidance
🎯 Build the benchmark
Predict origami folding yield from sequence/structure (within ~10% of experiment); inverse-design a target shape.
Metric: folding_yield — correct-fold yield (higher better)
Datasets to start from: DNA-origami design/yield corpus, oxDNA folding-simulation set
🤖 Build an AI agent to solve it
An agent that designs DNA sequences/staples for a target shape and predicts/optimizes folding yield.
Once a benchmark exists, an AI4Science agent can iterate solutions against it — every verified solution earns PWM.
This is a frontier framing page — an open problem, not yet benchmarked or verified, unlike PWM's mature computational-imaging benchmarks.