Under review as a conference paper at ICOMP 2024

Q UANTUM C OMPUTING IN D RUG D ISCOVERY: A R E -

VIEW OF Q UANTUM A NNEALING AND G ATE -BASED
A PPROACHES
Anonymous authors
Paper under double-blind review

A BSTRACT

Classical computers face significant challenges when dealing with NP problems,

especially given the unresolved question of whether NP equals P. These challenges
arise due to the computational complexity and resource limitations inherent in
solving such problems efficiently. Quantum computing, on the other hand, shows
promise in addressing these challenges through principles like superposition and
entanglement. By enabling parallel processing and potentially providing exponen-
tial speedup, quantum computing holds the potential to tackle problems that are in-
tractable for classical systems. One area where quantum advantage could be clear
is drug discovery. This review explores the role of quantum computing in drug
discovery, examining the state-of-the-art advancements in both quantum anneal-
ing and gate-based approaches. We discuss the current progress, the challenges
faced by these technologies, and provide recommendations for future research.
By identifying research gaps and potential areas for innovation, this review aims
to guide future advancements in applying quantum computing to drug discovery,
ultimately contributing to more effective and efficient methods for developing new
pharmaceuticals.

1 I NTRODUCTION
The trajectory of computing power, as predicted by Moore’s Law Schaller (1997), has long fueled
advancements in technology. However, as transistors shrink to nanoscale sizes, classical computers
encounter inherent limitations due to quantum effects Theis & Wong (2017). These limitations
arise from factors Kish (2004); Xiu (2019); Cavin et al. (2012) such as quantum tunneling and
interference, which become prominent at smaller scales. Quantum tunneling, for instance, leads
to electrons passing through barriers they would classically be unable to overcome, causing errors
and instability in computations. Additionally, quantum interference can disrupt the accuracy of
computations in classical systems. These quantum phenomena restrict the capacity of classical
computers to handle vast datasets and intricate simulations efficiently, highlighting the need for
alternative computing paradigms Shanbhag et al. (2008) to address these challenges effectively.
Actually, these quantum effects weren’t drawbacks Biswas et al. (2017); Córcoles et al. (2019);
Riel (2021). In fact, there was a groundbreaking suggestion by the well-known physicist Richard
Feynman to simulate complex quantum systems by using quantum mechanics itself to create com-
puters for doing this task Feynman (1960); Feynman et al. (1982); Feynman (1985). Feynman’s
insight, proposed in the early 1980s, laid the foundation for a new concept which is the quantum
computing, where quantum mechanics could be used to perform computations at a level far beyond
the capabilities of classical computers Khrennikov (2021). This shift in perspective from viewing
quantum effects as limitations to recognizing them as opportunities marked a pivotal moment in the
development of quantum computing.
Quantum superposition arises from Heisenberg’s uncertainty principle Hilgevoord & Uffink (2001);
Salloum et al. (2024a) , which states that the product of the uncertainties in position (∆x) and
momentum (∆p) of a particle is bounded by:
ℏ
∆x · ∆p ≥ (1)
2

1
Under review as a conference paper at ICOMP 2024

where ℏ is the reduced Planck constant (ℏ ≈ 1.054 × 10−34 Js).

This principle introduces inherent uncertainty into quantum systems, allowing a quantum bit (qubit)
to simultaneously represent both zero and one, a fundamental property exploited in quantum com-
puting.
The mathematical representation of superposition is given by Schrödinger’s equation McMahon
(2007):

|ψ⟩ = α|0⟩ + β|1⟩ (2)

Here, |ψ⟩ is the qubit state, α and β are complex probability amplitudes, and |0⟩ and |1⟩ represent
the zero and one states, respectively.
For example, consider a qubit in a superposition state:

1
|ψ⟩ = √ (|0⟩ + |1⟩) (3)
2

Here, α = √12 and β = √1 ,

2
representing equal probabilities of measuring zero or one when the
qubit is observed.
Entanglement Jozsa & Linden (2003) further enhances computational power in quantum systems.
By applying magnetic fields, entangled qubit systems can perform operations in a significantly
shorter time frame compared to classical systems. This is due to the correlated nature of entan-
gled qubits, allowing for efficient parallel computations.
The mathematical representation of entanglement is exemplified by the Bell state Horodecki et al.
(2009):

1
|Φ⟩ = √ (|00⟩ + |11⟩) (4)
2

The entangled state |Φ⟩ is a superposition of states |00⟩ and |11⟩, showcasing the unique correlations
created by entanglement.
This entangled state illustrates how measurements on one qubit instantaneously affect the state of
the other, demonstrating the efficiency of entanglement for quantum operations.
To understand the superposition and entanglement see the Figure 1.

Figure 1: Superposition and Entanglement. Adopted form Steptophysics (2023).

Additionally, a quantum system with n qubits can exhibit 2n states, showcasing the exponential
increase in computational possibilities in quantum computing.

2
Under review as a conference paper at ICOMP 2024

In the context of quantum computing paradigms, there are two primary models to consider Mizuno
& Yamaoka (2021): a quantum gate-based computer and a quantum annealer. Quantum annealing
Hauke et al. (2020); Salloum et al. is a quantum computing approach aimed at solving optimization
problems by finding the lowest energy state of a system, which corresponds to the optimal solution.
This process involves initializing a quantum system with the problem’s Hamiltonian and gradually
modifying it to guide the system towards the optimal solution. One notable leader in quantum an-
nealing is D-Wave Systems Koshka & Novotny (2020), which has demonstrated speedup in solving
combinatorial optimization problems using their quantum annealers Mott et al. (2017); Kumar et al.
(2018); Neukart et al. (2017); Neven et al. (2012); Von Dollen et al. (2021); Yarkoni et al. (2022).
In contrast, quantum gate-based computing Tiwari & Poonia (2021), also known as universal quan-
tum computing, utilizes quantum gates to manipulate qubits and perform quantum operations. This
approach allows for the execution of a wide range of quantum algorithms, including Shor’s algo-
rithm Shor (1994; 1999); Proos & Zalka (2003) for integer factorization and Grover’s algorithm
Grover (1996) for database search. Leaders in gate-based quantum computing include IBM Quan-
tum Alvarez-Rodriguez et al. (2018) and Google Quantum AI Courtland (2017).
The underlying principle of quantum annealing lies in leveraging quantum tunneling and super-
position to find the lowest energy state of a system, corresponding to the optimal solution of an
optimization problem Morita & Nishimori (2008). On the other hand, gate-based quantum comput-
ing uses quantum gates, such as Hadamard gates and CNOT gates, to manipulate qubits and perform
computations Michielsen et al. (2017). Gate-based quantum computing is versatile, capable of ex-
ecuting various quantum algorithms, and applicable across multiple problem domains, including
optimization, cryptography, simulation, and machine learning.
However, both approaches have their limitations. Quantum annealing is specialized for optimization
problems and may struggle with general-purpose quantum computing tasks beyond optimization. It
also has limited algorithmic flexibility compared to gate-based quantum computing. Connectivity
constraints in quantum annealers can restrict the types of optimization problems they can effectively
solve.
On the other hand, gate-based quantum computing faces challenges such as gate complexity, error
rates, and scalability issues. Implementing complex quantum algorithms on gate-based architectures
can be challenging, and gate errors, decoherence, and noise sources can limit the accuracy and
reliability of computations.
Therefore, quantum annealing and gate-based quantum computing offer distinct advantages and
limitations Karim Eddin et al. (2024), with quantum annealing excelling in optimization problems
and gate-based quantum computing showcasing versatility and potential speedup in various quantum
algorithms and problem domains. D-Wave Systems is a leader in quantum annealing, while IBM
Quantum and Google Quantum AI are leaders in gate-based quantum computing.
Both approaches can solve many problems that a classical computer may struggle with, spanning
various domains. One of the interesting and important domains where quantum computers can
provide significant help is drug discovery. Therefore, this work aims to provide a review of the
state-of-the-art (SOTA) for quantum applications in drug discovery Kumar et al. (2024) for both
quantum annealing and quantum gate-based computing.

S TRUCTURE OF THE PAPER

This paper is structured as follows: Section 2 provides an overview of drug discovery challenges,
while Section 3 discusses the state-of-the-art (SOTA) quantum annealing applications in drug dis-
covery. Following that, Section 4 covers the state-of-the-art (SOTA) quantum gate-based computing
applications in drug discovery. Furthermore, Section 5 presents the current advancements, potential,
and research gaps in quantum computing for drug discovery. Finally, Section 6 concludes the paper.

3
Under review as a conference paper at ICOMP 2024

Feature Quantum Annealing Gate-Based Quantum

Computing
Focus Optimization problems General-purpose quantum
computations
Technique Quantum tunneling and annealing Quantum gates and circuits
Strengths Specialized optimization Versatile, supports various
algorithms
Limitations Limited algorithm flexibility Gate complexity, error rates
Hardware Quantum annealers (e.g., D-Wave) Quantum processors (e.g., IBM,
Google)
Applications Combinatorial optimization, Cryptography, quantum simulation,
logistics, finance material science
Readiness Level Commercially available, mature Rapidly evolving, experimental
technology phase
Scalability Limited by coherence time and Potential for large-scale
noise computations with error correction
Algorithm Examples QUBO, Ising model Shor’s algorithm, Grover’s
algorithm
Implementation Difficulty Easier to implement for specific Requires complex quantum error
problems correction
Current Research Enhancing precision and scalability Error correction, fault-tolerant
quantum computing
Energy Efficiency Generally low due to specific Higher due to need for maintaining
hardware design coherence

Table 1: Comparison of Quantum Annealing and Gate-Based Quantum Computing

2 OVERVIEW OF C HALLENGES AND SOTA IN D RUG D ISCOVERY AND

C OMPUTATIONAL C HEMISTRY
The drug discovery process is a complex and multi-faceted endeavor that faces numerous challenges
across scientific, technical, regulatory, and ethical domains. These challenges significantly impact
the efficiency, cost, and success rates of bringing new drugs to market. Below is a comprehensive
overview of the key challenges encountered in drug discovery:

1. Complex Molecular Interactions: Understanding the interactions between drugs and bio-
logical targets Scott et al. (2016); He et al. (2010), such as proteins or enzymes, is inherently
complex due to the dynamic and intricate nature of molecular structures. Modeling these
interactions accurately often involves solving NP-hard problems, where finding the optimal
solution requires exponential time.
2. Complex Molecular Interactions: Understanding drug interactions with biological tar-
gets, such as proteins or enzymes Scott et al. (2016); He et al. (2010), is complex due
to the intricate nature of molecular structures. Accurately modeling these interactions
often involves NP-hard problems that require exponential time to solve. Quantum com-
puters have the potential to simulate molecular and atomic interactions with unparalleled
detail by using qubits that exist in superposition. This aligns with the quantum nature of
molecules, enabling more precise modeling. In drug discovery, quantum computing can
improve molecular docking accuracy by addressing quantum mechanical aspects, leading
to better predictions of binding affinities and identifying promising drug candidates.
3. High Dimensionality of Chemical Space: The vast chemical space containing potential
drug compounds presents a formidable challenge von Lilienfeld et al. (2020). Explor-
ing this space comprehensively to identify promising drug candidates is not only time-
consuming but also involves combinatorial optimization problems.

4
Under review as a conference paper at ICOMP 2024

4. Data Complexity: The vast amount of biological and chemical data available, including
genomics, proteomics, metabolomics, and structural data, poses challenges in data integra-
tion, analysis, and interpretation Grapov et al. (2018). Handling heterogeneous data sources
and extracting meaningful insights while managing data quality and biases is a significant
challenge.
5. Biological Complexity: Biological systems Hondermarck (2003) exhibit high levels of
complexity, including molecular interactions, cellular signaling pathways, and physiolog-
ical responses. Understanding the multifaceted nature of diseases and their underlying
biological mechanisms requires comprehensive modeling and simulation approaches.
6. Cost and Time Intensive: Developing a new drug from initial discovery to market ap-
proval is a lengthy and costly process Kiriiri et al. (2020); Petrova (2013), often taking up
to a decade and requiring substantial financial investments. The computational complexity
of analyzing large datasets and conducting extensive simulations further contributes to the
time and cost constraints.
7. Specificity and Efficacy: Ensuring that drugs are both specific to their intended targets
and efficacious in treating specific medical conditions is challenging. Designing drugs
with optimal therapeutic profiles requires solving optimization problems that are NP-hard
Nicolaou & Brown (2013), such as optimizing molecular structures for enhanced target
specificity and therapeutic efficacy.

In addressing these multifaceted challenges, emerging technologies such as quantum computing and
quantum machine learning offer promising avenues for innovation and optimization in drug discov-
ery processes. Leveraging the unique capabilities of quantum computing, including its potential to
solve NP-hard problems and process vast amounts of data, holds the potential to revolutionize drug
discovery workflows and accelerate the development of safe and effective treatments. Integrating
quantum machine learning algorithms into drug discovery pipelines can enhance predictive mod-
eling, target identification, and drug optimization processes, paving the way for more efficient and
personalized therapeutic interventions.

3 S TATE - OF - THE -A RT (SOTA) Q UANTUM A NNEALING A PPLICATIONS IN

D RUG D ISCOVERY

In this section, we review the state-of-the-art (SOTA) applications of quantum annealing in drug
discovery, highlighting its potential and current achievements.

1. Sampling Rare Protein Transitions Using Quantum Annealing: This study Ghamari
et al. (2024) demonstrates the potential of quantum annealing in simulating spontaneous
structural rearrangements in macromolecules, a task challenging for classical supercom-
puters due to time scale limitations. By employing a hybrid path-sampling paradigm that
combines classical exploration with quantum annealing for generating trial transition paths,
the study achieves significant progress in simulating complex protein transitions.
2. Designing Lattice Proteins with Quantum Annealing: Quantum annealing shows
promise in solving optimization problems related to protein folding and design Irbäck et al.
(2024). The study successfully identifies ground states in a coarse-grained lattice model
and optimizes sequences for protein structures, showcasing the effectiveness of quantum
annealing in biophysical challenges like protein design.
3. Ligand Modeling and Molecular Docking: Quantum annealing has been leveraged to
model ligand interactions with target proteins in drug discovery Shetty et al. (2023). By em-
ploying quantum-inspired algorithms combined with deep learning techniques, researchers
have achieved improvements in blind docking accuracy. This approach enables more effi-
cient exploration of molecular conformations and binding sites, leading to enhanced drug
design processes.
4. Molecular Unfolding and Docking Optimization: Quantum annealing has been applied
to optimize molecular docking processes, specifically in the phase of molecular unfold-
ing (SMU) Mato et al. (2022). By formulating the optimization problem as a High-order

5
Under review as a conference paper at ICOMP 2024

Unconstrained Binary Optimization (HUBO) and transforming it into a Quadratic Uncon-

strained Binary Optimization (QUBO), quantum annealing algorithms can efficiently ex-
plore the conformational space of molecules, improving docking accuracy and reliability.
5. Target Identification by Enzymes (TIE) Problem: Quantum optimization techniques
have been developed to address the NP-hard problem of target identification by enzymes
in metabolic networks. The QuTIE Ngo et al. (2023) (Quantum Optimization for Target
Identification by Enzymes) approach demonstrates optimal or near-optimal solutions for
identifying enzyme targets associated with specific diseases, showcasing the potential of
quantum annealing in complex biological network analysis.
6. Molecular Conformation Generation: Quantum-inspired algorithms have been utilized
to generate molecular conformations efficiently, a crucial step in structure-based drug de-
sign Li et al. (2024). By employing compact phase encoding methods and optimizing
internal distances within molecules, quantum-inspired approaches outperform traditional
methods like simulated annealing, offering faster and more accurate conformation genera-
tion for drug discovery purposes.
7. Machine Learning for Biomolecular Simulations: Quantum annealing has been explored
for machine learning tasks in computational biology Li et al. (2018), such as classifying
and ranking transcription factor binding affinities. While still in its early stages, quantum
machine learning approaches show promise in enhancing classification performance for
biological data sets, indicating a potential avenue for future advancements in drug discovery
research.
8. Multi-Objective Optimization in Drug Design: Quantum annealing approaches have
been applied to multi-objective optimization problems in drug design Tucs et al. (2023).
By considering multiple criteria, such as drug potency, selectivity, and ADMET (absorp-
tion, distribution, metabolism, excretion, and toxicity) Lin et al. (2003) properties, quantum
optimization techniques can guide the design of multi-functional and optimized drug can-
didates.

4 S TATE - OF - THE -A RT (SOTA) Q UANTUM G ATE -BASED C OMPUTING

A PPLICATIONS IN D RUG D ISCOVERY

Quantum gate-based approaches have shown tremendous promise in drug discovery, offering in-
novative solutions for molecular modeling, simulation, and analysis. Some key applications of
quantum gate-based methods in drug discovery include:

1. Quantum-Inspired Molecular Docking: Recent developments in quantum-inspired algo-

rithms, such as the digitized-counterdiabatic quantum approximate optimization algorithm
(DC-QAOA) Chandarana et al. (2022), have demonstrated remarkable accuracy in predict-
ing protein-ligand binding affinities Ding et al. (2024). These algorithms leverage quantum
principles to explore complex molecular interactions, offering unprecedented insights into
drug-target interactions crucial for designing highly specific and effective drugs.
2. Quantum-Based Biomolecular Simulations: Quantum computing enables high-fidelity
simulations of biomolecular systems Kim et al. (2022); Khatami et al. (2023), providing
detailed insights into molecular dynamics, conformational changes, and protein folding.
These simulations aid in understanding disease mechanisms, predicting drug efficacy, and
optimizing drug candidates for improved therapeutic outcomes.
3. Quantum Machine Learning (QML) for Drug Design: Quantum machine learning tech-
niques Batra et al. (2021); Vitz et al. (2024), including quantum-enhanced neural networks
and support vector machines, are being increasingly utilized for drug design and optimiza-
tion. These algorithms leverage quantum computing’s parallel processing capabilities to
analyze large datasets, identify molecular patterns, and predict pharmacological properties
with enhanced accuracy.
4. Hybrid Quantum-Classical Approaches: Hybrid quantum-classical algorithms, such as
the Variational Quantum Eigensolver (VQE) and Quantum Approximate Optimization Al-
gorithm (QAOA), offer scalable solutions for electronic structure calculations and energy

6
Under review as a conference paper at ICOMP 2024

minimization in drug molecules Innan et al. (2024). These approaches combine the compu-
tational power of quantum circuits with classical optimization techniques, achieving higher
accuracy and efficiency in modeling molecular properties.
5. Quantum-Assisted Virtual Screening: Quantum computing accelerates virtual screening
Mensa et al. (2023) processes by efficiently searching chemical databases and identifying
potential drug candidates with desired pharmacological profiles. Quantum-enhanced algo-
rithms streamline the screening pipeline, reducing time and resources required for identi-
fying lead compounds and accelerating drug discovery timelines.
6. Quantum-Based Biomolecular Structure Prediction: Gate-based quantum algorithms
facilitate biomolecular structure prediction, aiding in understanding protein folding mech-
anisms and ligand binding modes. These simulations contribute to the rational design of
therapeutically relevant molecules and optimize drug-target interactions Harris & Kendon
(2010); Wong & Chang (2022).
7. Quantum-Simulated Drug Design: Quantum simulations enable the design of novel drug
candidates by simulating chemical reactions, molecular transformations, and drug-target
interactions at the quantum level. These simulations facilitate rational drug design, opti-
mization of drug properties, and prediction of drug efficacy, leading to the development of
tailored therapeutics for diverse medical needs Andersson et al. (2022); Harris & Kendon
(2010).
8. Generative Design of Small Molecules: Quantum gate-based methods are employed in
generative design of small molecules, leveraging quantum algorithms to explore chemical
space and generate novel molecular structures with desired pharmacological properties.
These techniques enable the discovery of innovative drug candidates and accelerate drug
development pipelines. Examples include quantum generative adversarial networks Kao
et al. (2023) and quantum computing-enhanced algorithms Vakili et al. (2024).
9. Quantum Algorithms for Genomic Analysis: Quantum algorithms applied to genomic
data analysis expedite the identification of disease-related genetic variations, biomarkers,
and therapeutic targets Marchetti et al. (2022); Bhuvaneswari et al. (2023). Quantum-
enhanced genomic analysis tools enhance precision medicine initiatives by enabling per-
sonalized treatment strategies based on individual genetic profiles and disease mechanisms.

Quantum gate-based methods demonstrate significant potential in enhancing drug discovery work-
flows, from molecular docking and simulations to machine learning-driven drug design and virtual
screening.

5 C URRENT A DVANCEMENTS , P OTENTIAL , AND R ESEARCH G APS IN

Q UANTUM C OMPUTING FOR D RUG D ISCOVERY
While significant advancements have been made, it’s important to note that quantum computing
has not yet achieved widespread adoption or a major breakthrough in drug discovery. However,
the field shows immense potential for transformative impacts. Some notable progress includes im-
proved molecular modeling and simulation, enhanced drug design and optimization, accelerated
virtual screening, quantum machine learning for pharmacological predictions, precise biomolecular
structure prediction, quantum-assisted genomic analysis, and the development of hybrid quantum-
classical approaches. These advancements have contributed to a deeper understanding of complex
biological systems, faster identification of potential drug candidates, and more efficient drug design
processes.
However, several research gaps and challenges persist in harnessing quantum computing for drug
discovery:

1. Quantum Algorithm Optimization: Further optimization of quantum algorithms is

needed for specific drug discovery tasks, enhancing efficiency, accuracy, and scalability
for large-scale simulations.
2. Hardware Advancements: Improvements in quantum computing hardware are crucial to
overcome limitations such as qubit coherence, gate error rates, and scalability issues for
reliable computations.

7
Under review as a conference paper at ICOMP 2024

3. Data Integration and Validation: Developing frameworks for seamless integration of

diverse data sources into quantum platforms and establishing validation protocols against
experimental data are essential for reliable predictions.
4. Resource Optimization: Exploring methods to optimize resource utilization and reduce
computational costs in quantum simulations is necessary for practical applications.
5. Interpretability and Transparency: Ensuring clear explanations of quantum-derived re-
sults is crucial for gaining trust and acceptance in the pharmaceutical industry, especially
in machine learning-driven drug design.
6. Ethical and Regulatory Considerations: Addressing ethical and regulatory challenges
related to quantum computing in drug discovery is paramount for transparency, fairness,
and accountability.

By addressing these gaps, the intersection of quantum technology and pharmaceutical science can
achieve further exploration and innovation, leading to transformative advancements in drug discov-
ery and development.

6 C ONCLUSION AND O UTLOOK

Quantum computing presents a promising frontier in drug discovery, addressing challenges in
molecular modeling, optimization, and biomolecular simulations. Despite significant progress in
leveraging quantum technologies for tasks like ligand modeling and molecular docking, a consid-
erable gap remains between theoretical potential and practical implementation in the pharmaceuti-
cal industry. Advancements in quantum annealing systems, such as those from D-Wave Systems,
and gate-based quantum computing platforms like IBM Quantum and Google Quantum AI, en-
able researchers to explore complex molecular interactions, predict drug responses, and optimize
chemical reactions with unprecedented accuracy and efficiency. However, challenges persist, in-
cluding optimizing quantum algorithms for specific tasks, advancing quantum hardware, integrating
heterogeneous data sources, and ensuring interpretability and transparency. Ethical and regulatory
considerations also play a crucial role, necessitating frameworks for accountability and fairness.
In the short term, incremental improvements in quantum hardware and algorithms are expected,
leading to more practical applications in molecular modeling and simulation. The development of
specialized quantum algorithms may improve the accuracy and speed of drug candidate optimiza-
tion. Medium-term prospects include the integration of quantum and classical computing methods,
enhancing scalability and applicability to more complex biological systems and larger datasets. Col-
laborative efforts between academia, industry, and governmental bodies will be crucial in advancing
these technologies and addressing ethical and regulatory frameworks.
In the long term, as quantum computing technologies mature and become more accessible, they are
likely to lead to transformative changes in the pharmaceutical industry. Quantum computing could
become a standard tool in drug discovery, enabling highly targeted and personalized treatments.
Breakthroughs in quantum error correction and fault-tolerant quantum computing will enhance re-
liability and efficiency, solidifying their role in pharmaceutical research and development. The
future of quantum computing in drug discovery holds significant promise. By addressing current
challenges and fostering a collaborative ecosystem, the potential of quantum technologies to revo-
lutionize drug discovery and development is immense. This progression will ultimately contribute
to more effective, timely, and personalized therapeutic solutions, transforming modern medicine.
Continued research, collaboration, and innovation in this field will pave the way for groundbreaking
advancements in pharmaceutical science.

R EFERENCES
Unai Alvarez-Rodriguez, Mikel Sanz, Lucas Lamata, and Enrique Solano. Quantum artificial life in
an ibm quantum computer. Scientific reports, 8(1):14793, 2018.
Martin P Andersson, Mark N Jones, Kurt V Mikkelsen, Fengqi You, and Seyed Soheil Mansouri.
Quantum computing for chemical and biomolecular product design. Current Opinion in Chemical
Engineering, 36:100754, 2022.

8
Under review as a conference paper at ICOMP 2024

Kushal Batra, Kimberley M Zorn, Daniel H Foil, Eni Minerali, Victor O Gawriljuk, Thomas R Lane,
and Sean Ekins. Quantum machine learning algorithms for drug discovery applications. Journal
of chemical information and modeling, 61(6):2641–2647, 2021.
S Bhuvaneswari, R Deepakraj, Shabana Urooj, Neelam Sharma, and Nitish Pathak. Computational
analysis: unveiling the quantum algorithms for protein analysis and predictions. IEEE Access,
2023.
Rupak Biswas, Zhang Jiang, Kostya Kechezhi, Sergey Knysh, Salvatore Mandra, Bryan O’Gorman,
Alejandro Perdomo-Ortiz, Andre Petukhov, John Realpe-Gómez, Eleanor Rieffel, et al. A nasa
perspective on quantum computing: Opportunities and challenges. Parallel Computing, 64:81–
98, 2017.
Ralph K Cavin, Paolo Lugli, and Victor V Zhirnov. Science and engineering beyond moore’s law.
Proceedings of the IEEE, 100(Special Centennial Issue):1720–1749, 2012.
Pranav Chandarana, Narendra N Hegade, Koushik Paul, Francisco Albarrán-Arriagada, Enrique
Solano, Adolfo Del Campo, and Xi Chen. Digitized-counterdiabatic quantum approximate opti-
mization algorithm. Physical Review Research, 4(1):013141, 2022.
Antonio D Córcoles, Abhinav Kandala, Ali Javadi-Abhari, Douglas T McClure, Andrew W Cross,
Kristan Temme, Paul D Nation, Matthias Steffen, and Jay M Gambetta. Challenges and oppor-
tunities of near-term quantum computing systems. Proceedings of the IEEE, 108(8):1338–1352,
2019.
Rachel Courtland. Google aims for quantum computing supremacy [news]. IEEE Spectrum, 54(6):
9–10, 2017.
Arnab Das and Bikas K Chakrabarti. Quantum annealing and related optimization methods, volume
679. Springer Science & Business Media, 2005.
Qi-Ming Ding, Yi-Ming Huang, and Xiao Yuan. Molecular docking via quantum approximate
optimization algorithm. Physical Review Applied, 21(3):034036, 2024.
Richard P Feynman. There’s plenty of room at the bottom. Resonance, 16(9):890–905, 1960.
Richard P Feynman. Quantum mechanical computers. Optics news, 11(2):11–20, 1985.
Richard P Feynman et al. Simulating physics with computers. Int. j. Theor. phys, 21(6/7), 1982.
Danial Ghamari, Roberto Covino, and Pietro Faccioli. Sampling a rare protein transition using
quantum annealing. Journal of Chemical Theory and Computation, 2024.
Dmitry Grapov, Johannes Fahrmann, Kwanjeera Wanichthanarak, and Sakda Khoomrung. Rise of
deep learning for genomic, proteomic, and metabolomic data integration in precision medicine.
Omics: a journal of integrative biology, 22(10):630–636, 2018.
Lov K Grover. A fast quantum mechanical algorithm for database search. In Proceedings of the
twenty-eighth annual ACM symposium on Theory of computing, pp. 212–219, 1996.
Sarah A Harris and Vivien M Kendon. Quantum-assisted biomolecular modelling. Philosophi-
cal Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences, 368
(1924):3581–3592, 2010.
Philipp Hauke, Helmut G Katzgraber, Wolfgang Lechner, Hidetoshi Nishimori, and William D
Oliver. Perspectives of quantum annealing: Methods and implementations. Reports on Progress
in Physics, 83(5):054401, 2020.
Zhisong He, Jian Zhang, Xiao-He Shi, Le-Le Hu, Xiangyin Kong, Yu-Dong Cai, and Kuo-Chen
Chou. Predicting drug-target interaction networks based on functional groups and biological
features. PloS one, 5(3):e9603, 2010.
Itay Hen and Federico M Spedalieri. Quantum annealing for constrained optimization. Physical
Review Applied, 5(3):034007, 2016.

9
Under review as a conference paper at ICOMP 2024

Sovanmonynuth Heng, Dongmin Kim, Taekyung Kim, and Youngsun Han. How to solve combi-
natorial optimization problems using real quantum machines: A recent survey. IEEE Access, 10:
120106–120121, 2022.

Jan Hilgevoord and Jos Uffink. The uncertainty principle. 2001.

Hubert Hondermarck. Breast cancer: when proteomics challenges biological complexity. Molecular
& Cellular Proteomics, 2(5):281–291, 2003.

Ryszard Horodecki, Paweł Horodecki, Michał Horodecki, and Karol Horodecki. Quantum entan-
glement. Reviews of modern physics, 81(2):865, 2009.

Nouhaila Innan, Muhammad Al-Zafar Khan, and Mohamed Bennai. Quantum computing for elec-
tronic structure analysis: Ground state energy and molecular properties calculations. Materials
Today Communications, 38:107760, 2024.

Anders Irbäck, Lucas Knuthson, Sandipan Mohanty, and Carsten Peterson. Using quantum anneal-
ing to design lattice proteins. Physical Review Research, 6(1):013162, 2024.

Richard Jozsa and Noah Linden. On the role of entanglement in quantum-computational speed-up.
Proceedings of the Royal Society of London. Series A: Mathematical, Physical and Engineering
Sciences, 459(2036):2011–2032, 2003.

Po-Yu Kao, Ya-Chu Yang, Wei-Yin Chiang, Jen-Yueh Hsiao, Yudong Cao, Alex Aliper, Feng Ren,
Alán Aspuru-Guzik, Alex Zhavoronkov, Min-Hsiu Hsieh, et al. Exploring the advantages of
quantum generative adversarial networks in generative chemistry. Journal of Chemical Informa-
tion and Modeling, 63(11):3307–3318, 2023.

Suleiman Karim Eddin, Hadi Salloum, Mohamad Nour Shahin, Badee Salloum, Manuel Mazzara,
and Mohammad Reza Bahrami. Quantum microservices: transforming software architecture with
quantum computing. In International Conference on Advanced Information Networking and Ap-
plications, pp. 227–237. Springer, 2024.

Mohammad Hassan Khatami, Udson C Mendes, Nathan Wiebe, and Philip M Kim. Gate-based
quantum computing for protein design. PLOS Computational Biology, 19(4):e1011033, 2023.

Andrei Khrennikov. Roots of quantum computing supremacy: superposition, entanglement, or com-

plementarity? The European Physical Journal Special Topics, 230(4):1053–1057, 2021.

Philip Kim, Mohammad Hassan Khatami, Udson Mendes, and Nathan Wiebe. Gate-based quantum
computing for protein design. 2022.

Geoffrey Kabue Kiriiri, Peter Mbugua Njogu, and Alex Njoroge Mwangi. Exploring different ap-
proaches to improve the success of drug discovery and development projects: a review. Future
Journal of Pharmaceutical Sciences, 6:1–12, 2020.

LB Kish. Moore’s law and the energy requirement of computing versus performance. IEE
Proceedings-Circuits, Devices and Systems, 151(2):190–194, 2004.

Yaroslav Koshka and Mark A Novotny. Comparison of d-wave quantum annealing and classical sim-
ulated annealing for local minima determination. IEEE Journal on Selected Areas in Information
Theory, 1(2):515–525, 2020.

Gautam Kumar, Sahil Yadav, Aniruddha Mukherjee, Vikas Hassija, and Mohsen Guizani. Recent
advances in quantum computing for drug discovery and development. IEEE Access, 2024.

Vaibhaw Kumar, Gideon Bass, Casey Tomlin, and Joseph Dulny. Quantum annealing for combina-
torial clustering. Quantum Information Processing, 17:1–14, 2018.

Richard Y Li, Rosa Di Felice, Remo Rohs, and Daniel A Lidar. Quantum annealing versus clas-
sical machine learning applied to a simplified computational biology problem. NPJ quantum
information, 4(1):14, 2018.

10
Under review as a conference paper at ICOMP 2024

Yunting Li, Xiaopeng Cui, Zhaoping Xiong, Zuoheng Zou, Bowen Liu, Bi-Ying Wang, Runqiu Shu,
Huangjun Zhu, Nan Qiao, and Man-Hong Yung. Efficient molecular conformation generation
with quantum-inspired algorithm. arXiv preprint arXiv:2404.14101, 2024.
Jing Lin, Diana C Sahakian, SM De Morais, Jinghai J Xu, Robert J Polzer, and Steven M Winter.
The role of absorption, distribution, metabolism, excretion and toxicity in drug discovery. Current
topics in medicinal chemistry, 3(10):1125–1154, 2003.
Laura Marchetti, Riccardo Nifosı̀, Pier Luigi Martelli, Eleonora Da Pozzo, Valentina Cappello,
Francesco Banterle, Maria Letizia Trincavelli, Claudia Martini, and Massimo D’Elia. Quantum
computing algorithms: getting closer to critical problems in computational biology. Briefings in
Bioinformatics, 23(6):bbac437, 2022.
Kevin Mato, Riccardo Mengoni, Daniele Ottaviani, and Gianluca Palermo. Quantum molecular
unfolding. Quantum science and technology, 7(3):035020, 2022.
David McMahon. Quantum computing explained. John Wiley & Sons, 2007.
Stefano Mensa, Emre Sahin, Francesco Tacchino, Panagiotis Kl Barkoutsos, and Ivano Tavernelli.
Quantum machine learning framework for virtual screening in drug discovery: a prospective
quantum advantage. Machine Learning: Science and Technology, 4(1):015023, 2023.
Kristel Michielsen, Madita Nocon, Dennis Willsch, Fengping Jin, Thomas Lippert, and Hans
De Raedt. Benchmarking gate-based quantum computers. Computer Physics Communications,
220:44–55, 2017.
Hiroyuki Mizuno and Masanao Yamaoka. Research and development of annealing and gate-based
quantum computers. Hitachi Review, 70:486–490, 2021.
Satoshi Morita and Hidetoshi Nishimori. Mathematical foundation of quantum annealing. Journal
of Mathematical Physics, 49(12), 2008.
Alex Mott, Joshua Job, Jean-Roch Vlimant, Daniel Lidar, and Maria Spiropulu. Solving a higgs
optimization problem with quantum annealing for machine learning. Nature, 550(7676):375–
379, 2017.
Florian Neukart, Gabriele Compostella, Christian Seidel, David Von Dollen, Sheir Yarkoni, and Bob
Parney. Traffic flow optimization using a quantum annealer. Frontiers in ICT, 4:29, 2017.
Hartmut Neven, Vasil S Denchev, Geordie Rose, and William G Macready. Qboost: Large scale
classifier training withadiabatic quantum optimization. In Asian Conference on Machine Learn-
ing, pp. 333–348. PMLR, 2012.
Hoang M Ngo, My T Thai, and Tamer Kahveci. Qutie: quantum optimization for target identification
by enzymes. Bioinformatics Advances, 3(1):vbad112, 2023.
Christos A Nicolaou and Nathan Brown. Multi-objective optimization methods in drug design. Drug
Discovery Today: Technologies, 10(3):e427–e435, 2013.
Elina Petrova. Innovation in the pharmaceutical industry: The process of drug discovery and de-
velopment. In Innovation and Marketing in the Pharmaceutical Industry: Emerging Practices,
Research, and Policies, pp. 19–81. Springer, 2013.
John Proos and Christof Zalka. Shor’s discrete logarithm quantum algorithm for elliptic curves.
arXiv preprint quant-ph/0301141, 2003.
Atanu Rajak, Sei Suzuki, Amit Dutta, and Bikas K Chakrabarti. Quantum annealing: An overview.
Philosophical Transactions of the Royal Society A, 381(2241):20210417, 2023.
H Riel. Quantum computing technology. In 2021 IEEE International Electron Devices Meeting
(IEDM), pp. 1–3. IEEE, 2021.
Hadi Salloum, Ali Jnadi, Artem Matevosian, Sofia Gamershmidt, and Karam Almaghout. Enhanc-
ing robotic manipulation through fully integrated quantum annealing: The fiqa-rm system. In
International Conference on Computational Optimization.

11
Under review as a conference paper at ICOMP 2024

Hadi Salloum, Murhaf Alawir, Mohammad Anas Alatasi, Saleem Asekrea, Manuel Mazzara, and
Mohammad Reza Bahrami. Quantum advancements in securing networking infrastructures. In
International Conference on Advanced Information Networking and Applications, pp. 354–363.
Springer, 2024a.

Hadi Salloum, Hamza Shafee Aldaghstany, Osama Orabi, Ahmad Haidar, Mohammad Reza
Bahrami, and Manuel Mazzara. Integration of machine learning with quantum annealing. In
International Conference on Advanced Information Networking and Applications, pp. 338–348.
Springer, 2024b.

Robert R Schaller. Moore’s law: past, present and future. IEEE spectrum, 34(6):52–59, 1997.

Duncan E Scott, Andrew R Bayly, Chris Abell, and John Skidmore. Small molecules, big targets:
drug discovery faces the protein–protein interaction challenge. Nature Reviews Drug Discovery,
15(8):533–550, 2016.

Naresh R Shanbhag, Subhasish Mitra, Gustavode de Veciana, Michael Orshansky, Radu Marculescu,
Jaijeet Roychowdhury, Douglas Jones, and Jan M Rabaey. The search for alternative computa-
tional paradigms. IEEE Design & Test of Computers, 25(4):334–343, 2008.

Harshith Shetty, Aayush Joshi, Akshat Mehta, Rutuja Kumbhar, and Shahzia Sayyad. Leveraging
quantum annealing for ligand modelling in drug discovery. 2023.

Peter W Shor. Algorithms for quantum computation: discrete logarithms and factoring. In Proceed-
ings 35th annual symposium on foundations of computer science, pp. 124–134. Ieee, 1994.

Peter W Shor. Polynomial-time algorithms for prime factorization and discrete logarithms on a
quantum computer. SIAM review, 41(2):303–332, 1999.

Steptophysics. Superposition of qubits in quantum computing, 2023. URL https://www.

steptophysics.com/2023/07/superposition-of-qubits-in-quantum.
html.

Thomas N Theis and H-S Philip Wong. The end of moore’s law: A new beginning for information
technology. Computing in science & engineering, 19(2):41–50, 2017.

Yash Tiwari and Vishvendra Singh Poonia. Universal quantum gates based on quantum dots. arXiv
preprint arXiv:2105.07021, 2021.

Andrejs Tucs, Francois Berenger, Akiko Yumoto, Ryo Tamura, Takanori Uzawa, and Koji Tsuda.
Quantum annealing designs nonhemolytic antimicrobial peptides in a discrete latent space. ACS
Medicinal Chemistry Letters, 14(5):577–582, 2023.

Mohammad Ghazi Vakili, Christoph Gorgulla, AkshatKumar Nigam, Dmitry Bezrukov, Daniel
Varoli, Alex Aliper, Daniil Polykovsky, Krishna M Padmanabha Das, Jamie Snider, Anna Lyak-
isheva, et al. Quantum computing-enhanced algorithm unveils novel inhibitors for kras. arXiv
preprint arXiv:2402.08210, 2024.

Michael Vitz, Hamed Mohammadbagherpoor, Samarth Sandeep, Andrew Vlasic, Richard Padbury,
and Anh Pham. Hybrid quantum graph neural network for molecular property prediction. arXiv
preprint arXiv:2405.05205, 2024.

David Von Dollen, Florian Neukart, Daniel Weimer, and Thomas Bäck. Quantum-assisted feature
selection for vehicle price prediction modeling. arXiv preprint arXiv:2104.04049, 2021.

O Anatole von Lilienfeld, Klaus-Robert Müller, and Alexandre Tkatchenko. Exploring chemical
compound space with quantum-based machine learning. Nature Reviews Chemistry, 4(7):347–
358, 2020.

Renata Wong and Weng-Long Chang. Fast quantum algorithm for protein structure prediction in
hydrophobic-hydrophilic model. Journal of Parallel and Distributed Computing, 164:178–190,
2022.

12
Under review as a conference paper at ICOMP 2024

Liming Xiu. Time moore: Exploiting moore’s law from the perspective of time. IEEE Solid-State
Circuits Magazine, 11(1):39–55, 2019.
Sheir Yarkoni, Elena Raponi, Thomas Bäck, and Sebastian Schmitt. Quantum annealing for industry
applications: Introduction and review. Reports on Progress in Physics, 85(10):104001, 2022.
Lenny Putri Yulianti and Kridanto Surendro. Implementation of quantum annealing: A systematic
review. IEEE Access, 10:73156–73177, 2022.

A Q UANTUM G ATE - BASED AND Q UANTUM A NNEALING IN NUTSHELL

Quantum computing employs sophisticated methodologies such as Quantum Gate-based computing
and Quantum Annealing for advanced computational tasks.

A.1 Q UANTUM A NNEALING

Quantum Annealing Rajak et al. (2023); Hen & Spedalieri (2016); Salloum et al. (2024b); Yulianti
& Surendro (2022); Das & Chakrabarti (2005) tackles optimization problems by minimizing a cost
function E(s) using the annealing Hamiltonian H(s, t) = A(s) + B(t)C(s). The Ising model is
frequently employed in Quantum Annealing for tasks like graph partitioning and clustering.
X X
H=− Jij σiz σjz − hi σiz (5)
i,j i

Quantum Annealing techniques are pivotal in solving NP-hard problems and optimization chal-
lenges with exponential search spaces Rajak et al. (2023).
The Quantum Annealing process can be mathematically described using the Schrödinger equation
for a time-dependent Hamiltonian:

∂
iℏ |Ψ(t)⟩ = H(t)|Ψ(t)⟩ (6)
∂t
where |Ψ(t)⟩ is the quantum state at time t, ℏ is the reduced Planck constant, and H(t) is the time-
dependent Hamiltonian.
Quantum annealing is a metaheuristic tool considered a special and enhanced variant of simulated
annealing as shown in Figure 2. The core idea of simulated annealing is to probabilistically explore
different configurations of a system as the temperature decreases. The probability of transitioning
from one state to another depends on the energy difference between the states and is given by the
Boltzmann distribution:

e−βE
P (E) = , (7)
Z
where:
P (E) is the probability of the system being in state E,
1
β is the inverse temperature, β = , where kB is the Boltzmann constant and T is the temperature,
kB T
E is the energy of the state,
Z is the partition function, a normalization factor.

Quantum annealing, in contrast, operates with quantum states and relies on the Schrödinger equa-
tion:

H(t)|ψ(t)⟩ = E(t)|ψ(t)⟩, (8)

13
Under review as a conference paper at ICOMP 2024

Figure 2: Quatnum annealing vs simulated annealing. Adapted from Heng et al. (2022).

Figure 3: Quatnum annealing workflow. Adapted from Yarkoni et al. (2022).

where:
H(t) is the time-dependent Hamiltonian operator, |ψ(t)⟩ is the quantum state of the system at time t,
E(t) is the corresponding energy.

In drug discovery, quantum annealing has shown significant potential in addressing complex opti-
mization problems. By mapping the problem onto an Ising model, which is equivalent to a QUBO
(Quadratic Unconstrained Binary Optimization) problem, quantum annealing can efficiently explore
the vast search space to identify optimal configurations that minimize the system’s energy.
Algorithm 1 and Figure 3 outline and show the process of solving a problem using quantum anneal-
ing. First, the problem is defined and a cost function that quantifies the objective is constructed. The
cost function is then translated into a QUBO matrix, which is embedded onto a quantum annealer.

14
Under review as a conference paper at ICOMP 2024

The quantum annealer samples solutions to find the minimum energy states, which represent the
optimal solution to the problem.

A.2 Q UANTUM G ATE -BASED C OMPUTING IN A N UTSHELL

Quantum gate-based computing utilizes the unique properties of quantum gates to manipulate qubits,
enabling sophisticated computational processes that transcend classical computing capabilities. The
evolution of quantum states under the action of quantum gates is governed by unitary transforma-
tions, typically represented by the expression
U = e−iHt/ℏ
where H denotes the Hamiltonian operator, t represents time, and ℏ is the reduced Planck constant.
Key quantum gates such as the controlled-phase gate (CP ) and the Toffoli gate (CCN OT ) facilitate
complex multi-qubit operations, crucial for implementing quantum algorithms. The controlled-
phase gate (CP ) can be expressed as:
1 0 0 0
 
0 1 0 0 
CP = 0 0 1 0  ,
0 0 0 eiϕ
where ϕ is the phase angle. The Toffoli gate (CCN OT ), a three-qubit gate, is represented by:
1 0 0 0 0 0 0 0
 
 0 1 0 0 0 0 0 0
 0 0 1 0 0 0 0 0
 
 0 0 0 1 0 0 0 0
CCN OT =   0 0 0 0 1 0 0 0 .

 
 0 0 0 0 0 1 0 0
0 0 0 0 0 0 0 1
 
0 0 0 0 0 0 1 0
These gates are instrumental in the execution of pivotal quantum algorithms, such as the Quantum
Fourier Transform (QF T ) and the Quantum Phase Estimation (QP E). The QF T is essential for
decomposing quantum states into their frequency components, while QP E is utilized to estimate the
eigenvalues of unitary operators. These algorithms have profound implications for fields like cryp-
tography, where they underlie quantum factoring algorithms (e.g., Shor’s algorithm), and quantum
chemistry, where they facilitate the simulation of molecular systems.
In the realm of drug discovery, several quantum algorithms and circuits are of paramount impor-
tance due to their ability to simulate molecular interactions and predict chemical reactions with
high precision. These include the Variational Quantum Eigensolver (VQE), Quantum Approximate
Optimization Algorithm (QAOA), and Quantum Machine Learning Algorithms such as Quantum
Support Vector Machines (QSVM) and Quantum Neural Networks (QNN). These advanced quan-
tum computing methods pave the way for revolutionary advancements in computational science and
problem-solving paradigms. By utilizing these gates and circuits, researchers can achieve unprece-
dented accuracy in molecular simulations, accelerating the drug discovery process and enabling the
development of new therapeutics with greater efficiency and effectiveness.

B C ASE S TUDIES OF I NDUSTRY A PPLICATIONS OF Q UANTUM C OMPUTING

IN D RUG D ISCOVERY

In this section, we present several case studies that illustrate the application of quantum computing
technologies in the drug discovery industry. These case studies provide insights into how leading
companies are using quantum computing to address complex challenges in drug development.

B.1 D ISCOVERY OF N OVEL KRAS I NHIBITORS U SING A Q UANTUM

C OMPUTING -E NHANCED A LGORITHM

The discovery and development of small molecule inhibitors for KRAS, a crucial oncogene in cancer
therapy, have traditionally faced significant challenges. Researchers have begun to adopt innovative

15
Under review as a conference paper at ICOMP 2024

Algorithm 1 Quantum Annealing Pseudocode Example

1: function DefineProblem() {Define the problem’s variables and constraints.}
2: Problem.variables ← [”x1”, ”x2”, ”x3”]
3: Problem.constraints ← [”x1 + x2 ≤ 1”, ”x2 + x3 ≤ 1”]
4: return Problem
5: end function
6:
7: function CostFunction(Variables) {Create a cost function to minimize.}
8: Cost ← Variables[”x1”] + Variables[”x2”] + 2 * Variables[”x3”]
9: return Cost
10: end function
11:
12: function ConstructQUBOMatrix(Problem) {Translate cost function and constraints into a QUBO matrix.}

13: Q ← zeros matrix of appropriate size {Populate Q with coefficients from the cost function and con-
straints.}
14: Q[0][0] ← 1 {x1 coefficient}
15: Q[1][1] ← 1 {x2 coefficient}
16: Q[2][2] ← 2 {x3 coefficient}
17: Q[0][1] ← 2 {constraint: x1 + x2 ≤ 1}
18: Q[1][2] ← 2 {constraint: x2 + x3 ≤ 1}
19: return Q
20: end function
21:
22: function InitializeSampler() {Initialize the D-Wave sampler.}
23: sampler ← EmbeddingComposite(DWaveSampler())
24: return sampler
25: end function
26:
27: function EmbedProblem(Q, sampler) {Embed the QUBO matrix onto the quantum annealer.}
28: embedded problem ← sampler.sample qubo(Q)
29: return embedded problem
30: end function
31:
32: function SampleSolutions(embedded problem, sampler, num reads) {Sample solutions from the
quantum annealer.}
33: solutions ← sampler.sample(embedded problem, num reads)
34: return solutions
35: end function
36:
37: function CalculateEnergy(sample, Q) {Calculate the energy of a given sample based on the QUBO ma-
trix.}
38: energy ← 0
39: for (i, j), value ∈ Q do
40: if i ̸= bias & j ̸= bias then
41: energy ← energy + value · sample[i] · sample[j]
42: else if i == bias then
43: energy ← energy + value · sample[j]
44: else if j == bias then
45: energy ← energy + value · sample[i]
46: end if
47: end for
48: return energy
49: end function
50:
51: problem ← DefineProblem()
52: Q ← ConstructQUBOMatrix(problem)
53: sampler ← InitializeSampler()
54: embedded problem ← EmbedProblem(Q, sampler)
55: response ← SampleSolutions(embedded problem, sampler, 100)
56: min energy ← ∞
57: optimal sample ← None
58: for sample in response do
59: energy ← CalculateEnergy(sample, Q)
60: if energy < min energy then
61: min energy ← energy
62: optimal sample ← sample
63: end if 16
64: end for
65: Output: ”Optimal Sample: ”, optimal sample
Under review as a conference paper at ICOMP 2024

computational techniques to enhance the drug discovery process, aiming to increase hit rates and
decrease costs. This study introduced a hybrid quantum-classical generative model to design new
KRAS inhibitors, illustrated by a schematic representation of the hybrid framework for KRAS ligand
development Vakili et al. (2024).
In the initial phase, a curated set of 650 experimentally verified inhibitors targeting the KRAS protein
was extracted from the literature. The STONED-SELFIES algorithm was then applied to generate
analogs for each identified compound, expanding the collection to approximately 850,000 com-
pounds. An additional 250,000 top candidates were identified through a virtual screening process
using the REAL ligand library against the KRAS protein, culminating in a dataset of over 1 million
molecules for training the generative model.
The generative model training involved two key components: a classical Long Short-Term Memory
(LSTM) network and a Quantum Circuit Born Machine (QCBM). The LSTM network processed
sequential data encapsulating the chemical structures of ligands, while the QCBM, trained based
on the quality of LSTM-generated samples, created complex, high-dimensional probability distribu-
tions. Chemistry42 was used as a reward function to incentivize the creation of structurally diverse
and synthesizable molecules.
The workflow proceeded as follows:

1. Train the LSTM network on the compiled dataset.

2. Generate initial molecule candidates using the LSTM network.
3. Train the QCBM using the LSTM-generated samples.
4. Generate new molecules targeting KRAS using the QCBM.
5. Evaluate and refine the generated molecules using Chemistry42.

The pseudocode for this workflow is outlined below:

Algorithm 2 Hybrid Quantum-Classical Workflow for KRAS Ligand Development

1: Train LSTM network on training data
2: Generate initial molecules using LSTM model
3: Train QCBM on initial molecules
4: Generate quantum molecules using QCBM model
5: Initialize refined molecules as an empty list
6: for each molecule in quantum molecules do
7: score ← evaluate molecule using Chemistry42
8: if score > threshold then
9: add molecule to refined molecules
10: end if
11: end for
12: return refined molecules

Fifteen promising molecules were synthesized and subjected to experimental testing to assess their
ability to engage with the KRAS target. Among these candidates, two molecules, ISM061-018-2
and ISM061-22, demonstrated effective engagement with KRAS. ISM061-018-2 was identified as
a broad-spectrum KRAS inhibitor, exhibiting a binding affinity to KRAS-G12D at 1.4 µM, while
ISM061-22 exhibited specific mutant selectivity, displaying heightened activity against KRAS G12R
and Q61H mutants.
Comparative analysis with existing classical generative models indicated that integrating quantum
computing enhances distribution learning from established datasets, suggesting a potential advan-
tage for quantum generative models over their classical counterparts. The efficacy of distribution
learning was found to correlate with the number of qubits utilized, underlining the scalability poten-
tial of quantum computing resources.
This case study underscores the potential of quantum computing in enhancing drug discovery pro-
cesses. The integration of quantum algorithms with classical methods offers a promising approach
to developing effective therapeutics. The success of this study highlights the scalability potential of

17
Under review as a conference paper at ICOMP 2024

Figure 4: Representation of the KRAS Ligand Development Hybrid Framework. Adapted from
Vakili et al. (2024).

quantum computing resources, suggesting that future models could further improve drug discovery
efficiency and effectiveness.

B.2 H YBRID Q UANTUM G RAPH N EURAL N ETWORK FOR M OLECULAR P ROPERTY

P REDICTION

The integration of quantum computing and machine learning has led to the development of a hybrid
quantum-classical convoluted graph neural network (HyQCGNN) for predicting the formation ener-
gies of perovskite materials. This study utilized a gradient-free optimization approach and compared
the performance of HyQCGNN with classical models such as GENConv and XGBoost Vitz et al.
(2024).
The HyQCGNN model was implemented using PyTorch and PyTorch-Geometric for defining the
classical components and Qiskit for the quantum circuits. The GENConv layer performed dimen-
sionality reduction on the input graph, which was then processed by a quantum circuit using am-
plitude encoding. Observables from the quantum circuit were measured and compared to the target
values. The optimization was performed using the NGOpt algorithm from Nevergrad.
Training, evaluation, and software implementation details include:

18
Under review as a conference paper at ICOMP 2024

Algorithm 3 Hybrid Quantum-Classical Workflow for Molecular Property Prediction

1: Initialize optimizer with model parameters
2: Load classical parameters into GENConv layer
3: Transform input graph to intermediate graph using GENConv
4: Unroll intermediate graph into vector ⃗
x
5: Load ⃗ x into quantum circuit with quantum parameters
6: Measure observables, scale, and compare to target values
7: Report difference to optimizer for parameter update
8: Request updated parameters from optimizer

Figure 5: Performance of GENConv model

• The neural networks were defined using PyTorch and PyTorch-Geometric.

• Quantum circuits were defined using Qiskit and integrated with PyTorch through Qiskit’s
TorchConnector.
• Models were trained on an internal cluster with 64 CPU cores and simulated on the Qiskit
QASM simulator.
• The best models were selected based on performance on a validation set with 25 samples,
and final evaluation was done on a test set with 25 samples.

Both classical and hybrid models were trained for 2000 iterations using the NGOpt algorithm of
Nevergrad. The performance of HyQCGNN was competitive with the results obtained from GEN-
Conv and XGBoost models. Despite the advanced methods in GENConv, the hybrid model’s R2
score was only slightly lower, suggesting that hybrid quantum-classical models are viable for pre-
dicting complex molecular properties.
The results for both classical and hybrid methods are illustrated in the figures below:

• Figure 5: Plot and associated R2 value for the true formation energy vs. GENConv model
prediction.
• Figure 6: Plot and associated R2 value for the true formation energy vs. Hybrid model
prediction.
• Figure 7: Plot and associated R2 value for the true formation energy vs. XGBoost model
prediction.

The feature importance analysis performed using XGBoost highlighted that the most relevant feature
affecting the formation energy is the first ionization energy of site A in a perovskite material.

B.3 Q UANTUM A NNEALING IN M OLECULAR D OCKING

Drug discovery is a multi-phase process that includes virtual in silico simulations, in vitro, and
in vivo experimentation. Molecular docking, a critical step in this process, simulates the atomic
interactions of a ligand inside a protein binding site, predicting whether a stable complex can form.

19
Under review as a conference paper at ICOMP 2024

Figure 6: Performance of hybrid model

Figure 7: Performance of XGBoost model

This process involves significant computational resources due to the multiple degrees of freedom
and high dimensionality. Recent advancements in quantum computing have introduced potential
methods to enhance this computational process. Specifically, QA to support molecular docking,
focusing on the molecular unfolding (MU) process, which is the first step in geometric molecular
docking techniques Mato et al. (2022). MU aims to find the molecular configuration that maximizes
its volume by maximizing the internal distances between atoms within the molecule. The molecular
docking process involves detecting three-dimensional poses of the ligand within the active site of the
protein and ranking these poses using a scoring function. The initial ligand pose can introduce shape
bias, affecting docking quality. Molecular unfolding (MU) is used to remove this bias by expanding
the ligand to an unfolded shape. The MU problem’s objective is to maximize the molecular volume,
expressed as the total sum of internal distances between pairs of atoms in the ligand. Given a
molecule, the torsional configuration that maximizes this quantity is sought. The mathematical
formulation of the objective function for MU is given by:

X
D(t) = Dab (Θ)2 (9)
a,b∈M
a̸=b

where Dab (Θ) denotes the distance between two different atoms a and b within the molecule M ,
and Θ represents the set of torsional angles. Each torsion around a bond’s axis can assume values
within [0, 2π). The goal is to maximize D(t).
The process begins with identifying rotatable bonds, which are the problem parameters. These
discrete rotations are rewritten using one-hot encoding, introducing binary variables. The total sum
of internal atomic distances is expressed as a HUBO, transformed into a QUBO through distance
simplification, coarse-grained rotations, and threshold approximation. This reduction in complexity
enables embedding and running complex instances on QPUs.

20
Under review as a conference paper at ICOMP 2024

Algorithm 4 Molecular Unfolding Using Quantum Annealing

1: Identify rotatable bonds and initialize torsional angles.
2: Encode torsional angles using one-hot encoding to introduce binary variables.
3: Define the objective function D(t) as a HUBO.
4: Convert the HUBO to a QUBO.
5: Embed the QUBO on the quantum annealer.
6: Run quantum annealing and retrieve the solution.
7: Decode the solution to obtain the unfolded molecular configuration.

The quantum MU model was executed on D-Wave’s Advantage and 2000Q hardware. Performances
were compared with parallel random optimization, simulated annealing (SA), and the GeoDock
greedy algorithm. Advantage outperformed 2000Q in terms of qubits used and chain lengths. While
classical techniques excelled in larger problems, the quantum annealer yielded superior results for
medium-small problems, surpassing the GeoDock approach.
This study explored the potential of using a quantum annealer to enhance the drug discovery process,
specifically within the molecular unfolding phase of geometric molecular docking. By formulating
the MU problem as a high-order unconstrained binary optimization and solving it on D-Wave hard-
ware, we demonstrated the capabilities and limitations of current QA devices. Future work may
involve alternative encoding strategies, dynamic thresholds, and solution refinement techniques like
reverse annealing. Extending the QA approach to the entire docking process could further validate
the application of QC in computational sciences.

B.4 Q UANTUM C OMPUTING A PPLIED TO P ROTEIN F OLDING U SING THE L ATTICE -BASED
HP M ODEL

In the work Irbäck et al. (2024), the authors applied quantum computing technology to protein
folding using the lattice-based HP model. This model simplifies amino acids into hydrophobic
(H) and polar (P) groups. Previous studies on quantum computing methods for protein folding
have primarily relied on chain growth or turn-based algorithms. These approaches faced challenges
with nonlocal interactions unless the chain length was very short. Our novel approach employed
a scalable field-like representation with qubits at all lattice sites, enabling the study of chains up
to 64 amino acids long using the D-Wave hybrid solver. The work focused on the minimal two-
dimensional (2D) lattice-based HP model, where proteins are represented as self-avoiding chains of
H or P beads. These beads interact via a pairwise contact potential, and the energy function EHP is
defined as:
EHP = −NHH (10)

where NHH denotes the number of HH contacts. This model favors the formation of a hydropho-
bic core and serves as a useful test bed for novel computational approaches due to the availability
of exact results for all sequences with N ≤ 30. The design problem involves finding a sequence
s = (s1 , . . . , sN ) that folds into a given target structure Ct . the work aims to minimize the en-
ergy EHP (Ct , s) over sequence s using the D-Wave quantum annealer. This problem is recast in
quadratic unconstrained binary optimization (QUBO) form, with an auxiliary energy term to control
the number of H beads NH in the sequence. The total energy E(s) to minimize is given by:
N
!2
X X
E(s) = − wij si sj + λ si − NH (11)
1≤i<j≤N i=1

where si indicates whether bead i is of type P (si = 0) or H (si = 1). The parameter λ balances the
two terms to ensure the desired composition of H beads.
The D-Wave Advantage system provides a hybrid quantum-classical solver that combines classical
solvers with QPU queries to enhance performance on challenging QUBO problems. by utilizing this
solver to optimize sequences for target structures with N = 30, N = 50, and N = 64. For each
structure and composition, the authors performed multiple runs to generate optimized sequences.
Folding computations were then conducted using the D-Wave hybrid solver to verify that the se-
quences folded into the intended structures.

21
Under review as a conference paper at ICOMP 2024

The hybrid D-Wave annealer efficiently handled both steps of the sequence optimization and filtering
process, providing a robust approach to the HP design problem. Pure QPU computations were
limited by the problem size, showing decreased success rates for larger chains. Our time-dependent
Schrödinger equation simulations suggested that control errors in the Hamiltonian could explain this
limitation, as the added control noise qualitatively reproduced the modest QPU results.
The methods developed in this study are applicable to any quantum annealer and represent a sig-
nificant step towards practical applications of quantum annealing (QA) in protein design and other
biophysical challenges.

22