# Better Therapeutics with **Near-Term Quantum Computers**

*By Alireza Shabani and Anurag Mishra*

**Summary:** We posted a paper on ArXiv.org on the topic of protein simulation and quantum computing. We focused on the following question: What size of a quantum processor would be required to enhance the accuracy of protein simulation? Our research proposes that the number is a few hundred qubits, indicating that impactful pharmaceutical applications on near-term quantum computers is a realistic goal in the next three-five years.

Biomolecules such as proteins and DNA are the building blocks of the human body. Each cell in our body consists of millions of biomolecules with a wide range of sizes and structures that control all biological mechanisms. Understanding disease mechanisms at the molecular level has been the foundation of modern medicine. Cancer, for example, is caused by irregular behavior of biomolecules inside the patient’s cells that has proliferated to affect the whole body.

Since the emergence of transistor computers in the 1960s, digital simulation of biomolecule has been an active area of research at the intersection of chemistry and biology. In 2013, Levitt, Karplus, and Warshel won the chemistry Nobel prize for computational studies of functional properties of biological molecules. You may ask: why do we simulate biomolecules on a computer? That is because experimental measurements of biological systems, in vitro or in vivo, only tell us what the system is doing, and not how it is doing it. Therefore, it is crucial to simulate those systems on a computer in order to both understand experimental results and to make reliable predictions about how they function at the molecular level. It started as scientific curiosity half a century ago, now biomolecular simulations have become a reliable approach for discovery of novel therapeutics. We can design small molecule drugs on a computer because, we can compute how well they bind to a target protein, or we can decipher off-target effects in CRISPR by simulating DNA and RNA. Even with all these advances, however, there is still a long way to a fully-fledged computational approach to biology and therapeutics. Accuracy of simulations is a major bottleneck. We need accuracy at the quantum level.

What is quantum about biomolecules? When we look at a protein molecule, we observe a beautiful structure formed by a large group of tightly bound atoms. The various shapes of protein structures are dedicated by how different atoms bind to each other. If we zoom in, we observe cloud of electrons filling the space between atoms nuclei in particular spatial arrangements. This is where quantum physics comes into play. Molecules are quantum objects. The laws of quantum mechanics enables us to compute, with the precision of nature, how two atoms make a bond, then ten atoms, …, for a protein we need to simulate tens of thousands of interacting atoms!

Although in principle we can apply quantum laws to simulate natural systems, the size of a system that we can simulate on a computer is limited by the power of our classical computers. If one tries to store the information of a molecule’s electrons and orbitals in digital memory, the required size grows exponentially with the size of the simulation. So far we have managed to simulate larger and larger molecules by sacrificing the accuracy of the simulations, that has left as an open problem reliably simulating some basic systems such as interaction of water with biomolecules, or biomolecules that are electronically polarized.

A solution to the computational hardness of molecular simulations was put forward in 1982 by Dr. Richard Feynman: “Nature isn’t classical, dammit, and if you want to make a simulation of nature, you’d better make it quantum mechanical”. Basically, use a programmable quantum system to simulate another quantum system. Four decades later, we are indeed building a programmable quantum system, generally referred to as quantum computers.

Quantum computers utilize quantum effects, such as entanglement and superposition, to perform useful numerical computations. The most basic unit of information in quantum computing is the qubit (a quantum bit). A qubit is similar to a classical bit where it gives a result of 1 or 0 after a quantum algorithm is executed on it. Unlike a classical bit, a qubit can stay in superposition of both states, and can be *entangled* with another qubit such that an operation on one qubit simultaneously affects the result read from the second qubit. By clever use of superposition and entanglement, a quantum algorithm can provide a faithful and accurate simulation of a quantum system. In an article, we asked the question: “how many qubits do we need in order to leverage on quantum computers to improve the accuracy of protein simulation?” The numbers are promising.

At the heart of protein modeling is computing the interaction energies between different constituent atoms — how strongly they pull each other and how much freedom they have to jiggle around while being part of the same protein. In the context of drug design, we must compute how strong a drug molecule binds to a protein to validate it as a potential therapeutic. The common theory for biomolecule simulations model atoms nuclei as rigid balls applying forces on each other, referred to as force-fields. The force-field parameters are determined by quantum calculations and by fitting to experimental data for parts of a protein. Any protein molecule has a backbone made out of amino acids connected to each other via peptide bonds. Dipeptides are the smallest possible amino acid chain, consisting of two amino acids connected by a peptide bond. By learning the interaction of the two amino acid fragments in a dipeptide molecule, we can optimize a protein force field to accurately describe the interaction of many amino acid pairs with each other. Dipeptide molecules are large enough that their simulation is often done at a lower level of accuracy on classical computers. Quantum computers can potentially provide higher accuracy results with a limited number of qubits. In our paper, we estimated that with a minimum of 300 qubits one can reparametrize protein force-fields with more accurate quantum calculations performed on a Noisy Intermediate-Scale Quantum processors.

We expect that quantum processors with hundreds of qubits to become practical over the next 3–5 years. Thus, our results make it an achievable goal to have impactful pharmaceutical applications for near-term quantum processors.