Quantum PCA for Gene-Expression Data

Gene-Expression Cohort

Correlation ρ across genes 0.60

Cohort size N 160

Classical PCA (power iteration)

λ₁ (PC1)

–

λ₂ (PC2)

–

λ₃ (PC3)

–

PC1 variance

–

Quantum Phase Estimation on ρ

Ancilla (precision) qubits t 4

φ target = 0.000 → estimate = 0.000

bin 0bin 15

Classical reads

–

Quantum qubits

–

How it works

Each point is a "sample" in a 3-gene expression space. Classically, PCA power-iterates the empirical covariance matrix C = (1/N)Σ(xᵢ−x̄)(xᵢ−x̄)ᵀ:

v ← Cv / |Cv|,  λ ← vᵀCv   (repeat to converge)

The arrows in the 3D view are the converged eigenvectors, scaled by √λ. Quantum PCA instead treats the (normalized) covariance matrix as a density operator ρ = C/tr(C) and exponentiates many copies of ρ to synthesize e^(−iρt); running Quantum Phase Estimation (QPE) on that unitary extracts each eigenvalue λ using only t = O(log(1/ε)) ancilla qubits, versus reading every one of the d² classical matrix entries.

The histogram below is the exact QPE outcome distribution for measuring the top eigenvalue's phase φ = λ₁ / (2·tr C) with t ancilla qubits — the real Nielsen–Chuang result, not a mock-up:

P(y) = sin²(π(2ᵗφ − y)) / (2²ᵗ · sin²(π(φ − y/2ᵗ)))

Raise t and watch the peak sharpen around the true phase: that narrowing is the source of QPCA's claimed exponential advantage for high-dimensional genomic covariance matrices — at the cost of needing coherent access to many copies of the quantum state, which is why it remains a NISQ-era research idea rather than a deployed bioinformatics tool.

Expression samples Principal axes (√λ scaled)

⚠ Couldn't load the 3D engineThree.js failed to load from the CDN. Check your connection and reload.