⚛️ quantum physics

Quantum feedback algorithms for DNA assembly using FALQON variants

This paper demonstrates that Feedback-based Algorithm variants (FALQON, SO-FALQON, and TR-FALQON) effectively improve convergence and success probabilities for de novo DNA assembly on near-term quantum hardware by eliminating classical optimization loops through measurement feedback.

Original authors: Pedro M. Prado, Lucas A. M. Rattighieri, Rafael Simões do Carmo, Giovanni S. Franco, Guilherme E. L. Pexe, Alexandre Drinko, Erick G. Dorlass, Tatiana F. de Almeida, Felipe F. Fanchini

Published 2026-02-25

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Pedro M. Prado, Lucas A. M. Rattighieri, Rafael Simões do Carmo, Giovanni S. Franco, Guilherme E. L. Pexe, Alexandre Drinko, Erick G. Dorlass, Tatiana F. de Almeida, Felipe F. Fanchini

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you have just shredded a massive, complex instruction manual (like a user guide for a human body or a virus) into thousands of tiny, overlapping pieces of paper. Your goal? To tape them back together to reconstruct the original manual, but you don't have the original to compare it against. This is the challenge of de novo DNA assembly.

In the real world, computers try to solve this by looking for overlapping edges between the paper scraps. However, when the manual is huge and the pieces are messy (due to errors or repeated phrases), it becomes a nightmare for classical computers. It's like trying to solve a jigsaw puzzle where every piece looks slightly similar to its neighbors, and you have billions of pieces.

This paper proposes a new way to solve this puzzle using quantum computers, specifically using a "smart feedback" system called FALQON and its two upgraded cousins.

Here is the breakdown of their approach using simple analogies:

1. The Problem: The "Noisy" Jigsaw Puzzle

The authors take DNA fragments (like from the SARS-CoV-2 virus or human mitochondria) and turn the assembly problem into a math game called QUBO.

The Analogy: Imagine you are trying to find the perfect seating arrangement for a dinner party. Everyone has specific preferences about who they want to sit next to (overlaps) and who they don't want to sit next to (errors). You need to find the one arrangement where everyone is happy.
The Issue: Traditional quantum algorithms (like QAOA) are like a student trying to solve this by guessing a seating chart, asking a teacher (a classical computer) "Is this better?", getting an answer, and then guessing again. This "guess-and-check" loop is slow, gets confused by noise (static), and often gets stuck in a bad arrangement.

2. The Solution: The "Self-Correcting" Robot

The authors use an algorithm called FALQON.

The Analogy: Instead of asking a teacher for help, imagine a robot that can feel the table. If the table is wobbly (high energy/bad arrangement), the robot instantly knows which way to push the chairs to make it stable. It doesn't need to stop and think; it just reacts to the physics of the situation.
How it works: The quantum computer measures the "energy" of the current DNA arrangement. If the energy is high (bad assembly), the algorithm automatically adjusts the quantum circuit to lower the energy. It's a continuous, self-correcting flow that eliminates the need for the slow "guess-and-check" loop.

3. The Upgrades: Making the Robot Faster and Smarter

While the original FALQON robot works, it can be a bit slow or take too many steps to reach the perfect solution. The paper introduces two "turbo-charged" versions:

A. TR-FALQON (Time-Rescaled FALQON)

The Analogy: Imagine driving a car toward a destination. The standard robot drives at a constant, cautious speed. TR-FALQON is like a driver who knows the road ahead. It speeds up on the straight, empty highways (where the solution is easy to find) and slows down only when it hits a tricky curve (a complex part of the DNA).
The Result: It gets to the destination (the correct DNA sequence) much faster and with fewer turns (fewer circuit layers), which is crucial because current quantum computers are fragile and can't handle long, complex drives.

B. SO-FALQON (Second-Order FALQON)

The Analogy: The standard robot looks at the immediate slope of the hill to decide which way to push. SO-FALQON is like a robot that looks at the slope and how steep the slope is changing (the curvature). It predicts the future path more accurately.
The Result: It can take bigger, bolder steps without falling off the cliff. This means it reaches the solution in fewer steps, making it much more efficient for noisy, imperfect quantum hardware.

4. The Results: Winning the Race

The authors tested these three robots (Standard, Time-Rescaled, and Second-Order) on real DNA data from a virus and human mitochondria.

The Outcome: The upgraded robots (TR and SO) were significantly better. They found the correct DNA sequence with higher accuracy and, most importantly, did it using fewer steps.
Why this matters: Current quantum computers are like "noisy" devices that break down if you ask them to do too many steps at once. By finding the solution in fewer steps, these new algorithms make it possible to solve complex biological problems on the quantum computers we have today, rather than waiting for perfect machines in the future.

The Bottom Line

This paper is about teaching quantum computers to be better at "feeling" their way through complex biological puzzles. By removing the need for slow, external thinking and adding "smart" speed adjustments and predictive steps, they have created a method that is faster, more accurate, and ready to work on the imperfect quantum hardware available right now. It's a major step toward using quantum computers to revolutionize medicine and biology.

1. Problem Statement

The paper addresses the computational challenge of de novo DNA assembly, which involves reconstructing a complete genomic sequence from short, overlapping sequencing reads without a pre-existing reference genome.

Complexity: The problem is complicated by genomic repetitions, sequencing errors, and the sheer volume of data.
Current Limitations: Classical solutions (e.g., de Bruijn graphs, Velvet, SPAdes) face scalability and accuracy issues in large genomes.
Quantum Formulation: The authors map the assembly problem to a Quadratic Unconstrained Binary Optimization (QUBO) problem, which is equivalent to finding the ground state of an Ising Hamiltonian. This formulation treats the assembly as a Traveling Salesman Problem (TSP) on an overlap graph, where the goal is to find a Hamiltonian cycle that maximizes the sum of overlaps between reads.
Hardware Constraints: Standard variational quantum algorithms (like QAOA) rely on classical optimization loops to tune circuit parameters. In the Noisy Intermediate-Scale Quantum (NISQ) era, these loops are prone to noise, barren plateaus (vanishing gradients), and high computational costs, limiting the circuit depth and problem size.

2. Methodology

The authors propose and evaluate three quantum algorithms based on Lyapunov control, which eliminate the need for classical parameter optimization by using measurement feedback to dynamically adjust circuit parameters.

A. The Base Algorithm: FALQON

Mechanism: FALQON (Feedback-based Algorithm for Quantum Optimization) uses the expectation value of the commutator between the problem Hamiltonian ( $H_p$ ) and a driver Hamiltonian ( $H_d$ ) to update the control parameter $\beta_k$ at each layer.
Feedback Law: $\beta_{k+1} = -i\langle\psi_k|[H_d, H_p]|\psi_k\rangle$ .
Goal: This ensures a monotonic descent of the expected energy, driving the system toward the ground state without a classical optimizer.

B. Variant 1: Time-Rescaled FALQON (TR-FALQON)

Concept: Introduces a temporal rescaling function $f(\tau)$ to redefine the evolution time.
Mechanism: The evolution operator is modified by a scaling factor $\dot{f}(\tau)$ . When $\dot{f}(\tau) > 1$ , evolution is accelerated; when $< 1$ , it is slowed down to allow higher resolution in critical regions.
Benefit: Aims to accelerate convergence and achieve high fidelity at shallower circuit depths, which is crucial for NISQ devices with limited coherence times.

C. Variant 2: Second-Order FALQON (SO-FALQON)

Concept: Improves the feedback law by including a second-order Taylor expansion of the evolution operators.
Mechanism: The update rule for $\beta_k$ incorporates quadratic terms involving double commutators ( $[[H_d, H_p], H_d]$ and $[[H_d, H_p], H_p]$ ).
Adaptive Strategy: The algorithm adaptively chooses between the first-order and second-order updates based on the magnitude of the terms to ensure numerical stability and maintain the condition $\Delta\langle H_p \rangle < 0$ .
Benefit: Allows for larger time steps and a significant reduction in the total circuit depth required for convergence.

D. Implementation Details

Dataset: Experiments used long-read DNA fragments from SARS-CoV-2 and human mitochondrial DNA.
Encoding: The QUBO formulation uses a "one-hot" encoding where binary variables $x_{i,p}$ indicate if read $i$ is at position $p$ . The problem was mapped to an Ising Hamiltonian using Pauli $Z$ operators.
Qubit Count: For $n$ reads, the encoding requires $(n-1)^2$ qubits (fixing the first read to reduce qubit count). Experiments ranged from 4 to 6 reads (9 to 25 qubits).

3. Key Contributions

Algorithmic Innovation: The first application of TR-FALQON and SO-FALQON variants specifically to the combinatorial problem of DNA assembly.
Hybrid-Free Optimization: Demonstrated a fully quantum feedback loop that removes the classical optimization bottleneck, making the approach more robust against noise and barren plateaus.
Performance Analysis: Provided a rigorous comparison showing that the variants outperform standard FALQON in terms of convergence speed and success probability, particularly under depth constraints.
Biological Benchmarking: Validated the methods on real-world biological data (SARS-CoV-2 and human mitochondrial DNA), proving the feasibility of quantum approaches for bioinformatics tasks.

4. Results

The authors conducted numerical simulations comparing FALQON, TR-FALQON, and SO-FALQON across different hyperparameter sets and read counts (4, 5, and 6 reads).

Convergence: Both TR-FALQON and SO-FALQON demonstrated faster convergence to the ground state energy compared to standard FALQON.
Circuit Depth: The variants achieved high success probabilities at reduced circuit depths.
- TR-FALQON: Achieved the lowest final energy and the highest success probability (>50% in some configurations) with fewer layers.
- SO-FALQON: Showed robustness even with larger time intervals, maintaining stability where standard FALQON might stagnate.
Scalability: As the number of reads increased (expanding the Hilbert space dimension from $2^9$ to $2^{25}$ ), the advantage of the enhanced feedback strategies became more pronounced.
Success Probability: In the SARS-CoV-2 5-read instance, TR-FALQON variants saturated at success probabilities around 40–50%, significantly outperforming the standard FALQON which lagged behind.

5. Significance and Conclusion

This work highlights that enhanced feedback-driven dynamics are a viable and superior strategy for solving combinatorial optimization problems on near-term quantum hardware.

NISQ Relevance: By reducing the required circuit depth, TR-FALQON and SO-FALQON mitigate the effects of noise and decoherence, making them suitable for current quantum processors.
Bioinformatics Impact: The study opens a pathway for applying quantum computing to complex genomic reconstruction tasks, potentially reducing research costs and enabling real-time analysis in clinical diagnostics.
Future Outlook: The authors suggest that these feedback-based approaches represent a promising direction for scaling quantum algorithms to larger, more complex genome assembly problems and other bioinformatics challenges.

In summary, the paper demonstrates that moving beyond first-order feedback (standard FALQON) to include temporal rescaling and second-order approximations significantly enhances the efficiency and robustness of quantum algorithms for DNA assembly.