Use the Dynex platform to find the optimal stem structure for HIV and Tobacco Mosaic Virus RNA sequences. This example takes an RNA sequence and applies a quadratic model to pursue the optimal stem configuration.
Refer to the original dynex use case:
https://github.com/dynexcoin/DynexSDK/blob/main/example_rna_folding.ipynb
Quantum Algorithm Contributors:
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9022793/pdf/pcbi.1010032.pdf
Knowledge background: This is a quantum computer algorithm
Not all programs running on quantum computers have the so-called "quantum advantage". Many of them can actually be simulated by classical computers. These methods of efficiently simulating quantum computing processes have spawned a number of "quantum inspired algorithms". For example, the simulated bifurcation algorithm was derived from the quantum adiabatic theorem. This is an algorithm running on classical computers and has been very popular in the field of combinatorial optimization in recent years.
However, there are some quantum computing processes that are difficult to simulate with classical computers (such as Shor's algorithm). One of the key features is the use of quantum entanglement. In the field of condensed matter physics, people have found that it is very difficult to simulate "quantum many-body systems" with strong entanglement on classical computers. Currently, the more effective one is the tensor network algorithm, but it is also limited by the scale of the problem. It is generally believed that it is difficult to simulate systems with strong entanglement with classical computers.
Quantum entanglement content assessment
The 3-dimensional folding of RNA molecules depends strongly on the pattern of intramolecular hydrogen bonds between bases. Predicting the base pairing network from a sequence, also known as RNA secondary structure prediction or RNA folding, is a non-deterministic polynomial time (NP) complete computational problem. The structure of a molecule can strongly predict its functional and biochemical properties, so the ability to accurately predict structure is an important tool for biochemists. Many methods have been proposed to efficiently sample possible secondary structure patterns. Classical methods employ dynamic programming, and recent studies have explored methods inspired by evolutionary and machine learning algorithms. This work demonstrates the use of quantum computing hardware to predict the secondary structure of RNA. A Hamiltonian written in the form of a binary quadratic model (BQM) is derived to drive the system to maximize the number of consecutive base pairs while jointly maximizing the average length of the stem. The quantum annealer (QA) is compared to the replica exchange Monte Carlo (REMC) algorithm programmed with the same objective function, and the results show that QA is highly competitive in quickly identifying low-energy solutions.
Quantum entanglement detection
To date, there are two main models of quantum computers (QC): gate model and quantum annealing. Gate model quantum devices have a wide range of applications and are most commonly used in quantum chemistry and quantum machine learning computations. Another design pioneered by D-Wave is QA. Compared to the many applications of gate model QC, QA is much narrower in scope and focuses solely on optimizing the solution to a problem by minimizing the problem Hamiltonian. To date, QAs containing thousands of qubits have been built, and these devices are capable of solving discrete combinatorial optimization problems large enough to allow testing for real-world industrial use cases. Instead of being programmed by a sequence of quantum operators, QAs are designed to anneal a quadratic Hamiltonian in the form of an equation.
The D-Wave QA used in the prototype for this study is an analog device containing about 5,000 qubits. The device is programmed by setting the values of the local magnetic field and the coupling strength, and the annealing process is performed by adiabatically reducing the strength of the transverse magnetic field. The design is similar to simulated annealing of the Ising Hamiltonian, with the main differences being that the annealing process avoids getting stuck in the local minimum through quantum tunneling rather than thermal fluctuations, and the probability of jumping out of the local minimum is determined by the width of the barrier rather than its height.
Current methods for performing computations on quantum devices require that the interaction terms be pre-computed on the classical device and read into the quantum device via a specialized API. The program was built and executed using the python library provided by D-Wave Systems. The BinaryQuadraticModel class from the dimod 0.9.10 python library was used to build the model from classically prepared data and convert it into a data structure compatible with the quantum device. The one- and two-body interaction terms were pre-computed, stored in numpy arrays, and passed into the BinaryQuadraticModel instance as a dimod.BINARY representation along with an offset of 0.0. The model was executed using the LeapHybridSampler class from the dwave.system python library. An execution time of 3 seconds was allotted to the solver. The eigenstate with the lowest correlation eigenvalue was selected to represent the simulation results.
This study migrated to Dynex
The Dynex SDK provides a neuromorphic/QUBO sampler that can be called from any Python code. Developers and application developers already familiar with the DIMOD framework will find it very easy to run calculations on the Dynex neuromorphic computing platform: the Dynex Sampler object can simply replace the default sampler object that is usually used to run calculations, such as D-wave systems, without the limitations of quantum machines.
