Electron transfer in biological systemsResearch 1

Using a combination of atomistic molecular dynamics (MD) simulations and density functional theory (DFT) electronic structure calculations, we investigate electron transfer mechanisms and structure-function correlations in nucleic acids, self-standing or interacting with proteins. 

  • S. Nandi, B. Rudd, R. Di Felice, “Axial electronic conduction channels in duplex DNA”, manuscript under review. 
  • S. Nandi, C. Coane, A.-E. Villegas, A. Ray, R. Di Felice, “The impact of G-quadruplex dynamics on inter-tetrad electronic couplings: a hybrid computational study”, Phys. Chem. Chem. Phys. 24, 22513 (2022). 
  • M. Mantela, C. Simserides, R. Di Felice, “LCAO electronic structure of nucleic acid bases and other heterocycles and transfer integrals in B-DNA, including structural variability”, Materials 14, 4930 (2021). 
  • M. Mantela, A. Morphis, K. Lambropoulos, C. Simserides, R. Di Felice, “Effects of structural dynamics on charge carrier transfer in B-DNA: a combined MD and RT-TDDFT study”, J. Phys. Chem. B 125, 3986 (2021). 
  • R. D. Teo, B. J. G. Rousseau, E. R. Smithwick, R. Di Felice, D. N. Beratan, A. Migliore, “Delineating the mechanisms of DNA-mediated signaling between [4Fe4S] proteins”, Chem. 5, 1-16 (2019). 

Molecular simulations Research 2

We investigate protein-DNA complexes by means of atomistic MD simulations over multi-microsecond time scales. Examples of our studies include ternary DNA-bound CRISPR/Cas9, in which we have inspected the role of protein mutation in structural conformations that may affect editing specificity. In transcription-factor/DNA complexes, we look at binding-induced DNA conformational changes that may reveal mechanisms for binding specificity, accompanied by experimental data analysis with machine learning and more. 

  • S.H. Fatemi Abhari, S. Nandi, A. Ray, R. Di Felice, “Molecular Dynamics simulations of DNA-bound CRISPR/Cas9 complexes on different time scales”, manuscript under review. 
  • S.H. Fatemi Abhari and R. Di Felice, “Probing electrostatic interactions in DNA-bound CRISPR/Cas9 complexes by molecular dynamics simulations”, ACS Omega 9, 44974–44988 (2024). 
  • S. Lara-Gonzalez, A. C. Dantas Machado, S. Rao, A. A. Napoli, J. Birktoft, R. Di Felice, R. Rohs, C. L. Lawson, “The RNA Polymerase a Subunit Recognizes the DNA Shape of the Upstream Promoter Element”, Biochemistry 59, 4523−4532 (2020). 
  • A. C. Dantas Machado, B. H. Cooper, X. Lei, R. Di Felice, L. Chen, R. Rohs, “Landscape of DNA binding signatures of myocyte enhancer factor-2B reveals a unique interplay of base and shape readout”, Nucleic Acids Research 48, 8529-8544 (2020). 
  • A. Ray and R. Di Felice, “Protein-Mutation-Induced Conformational Changes of the DNA and Nuclease Domain in CRISPR/Cas9 Systems by Molecular Dynamics Simulations”, J. Phys. Chem. B 124, 2168 (2020). 
  • A. Ray and R. Di Felice, “Molecular simulations have boosted knowledge of CRISPR/Cas9: A Review”, Journal of Self-Assembly and Molecular Electronics 7, 45 (2019).  

Quantum computational (cancer) drug discoveryresearch #3

Computational drug discovery relies on a variety of computational tasks, including the creation of virtual chemical spaces and libraries (putative drugs) and the screening of such libraries for potential interaction with target molecules (proteins or nucleic acids). These tasks aim at the quantification of bond strengths, which are in the realm of quantum mechanics. The bonds that need to be quantified are those between atoms in a molecule, which bring from the chemical formula (e.g., C16H18N2O4S for penicillin) to the molecular geometry, and those between a given molecular (drug) structure and a target protein, which eventually determine if a drug is effective and specific.  The complexity of the quantum mechanical problem increases exponentially with system size. The direct consequence is that an accurate prediction of the bond strengths is intractable for the systems of interest in medicine. Thus, researchers need to screen colossal chemical libraries to avoid the omission of strong binders in the pipeline: the number of organic compounds that could be synthesized is enormous, routinely cited as greater than a quadrillion of quadrillions. While the development of artificial intelligence tools may enable this formidable screening task across a wide range of medical scenarios, the process still produces a large number of compounds to be tested in laboratories. The accurate knowledge of bond strengths could revolutionize the field, by ideally yielding a single drug candidate, or few. It can work in synergy with current AI protocols. For their nominal linear scaling with system size to solve quantum mechanical problems with potentially 100% accuracy, quantum algorithms executed on quantum processing units promise to make a transformative impact on computational drug design. In this work, we propose an end-to-end computational pipeline that includes: (1) MD and QM/MM simulations to identify the portion of the system that necessitates a correlated electronic structure description; (2) active space reduction and quantum embedding to produce a manageable molecular Hamiltonian; (3) the development of a quantum algorithm for ground-state energy estimation and its demonstration on quantum simulators and quantum hardware. We develop the pipeline on the test case of cancer drugs.

This work is in progress for the NIH quantum computing challenge 2024: https://www.nih.gov/challenges/nih-quantum-computing-challenge. 

A quantum computing protocol for corrosion inhibition and heterogeneous catalysisresearch 4

We developed a hybrid classical-quantum computational pipeline for the determination of adsorption energies of a benzotriazole molecule on an aluminum alloy surface relevant for the transport industry, to address the corrosion problem. The molecular adsorbate and substrate alloy were selected by interrogating molecular and materials databases, in search for desired criteria. The protocol can be generalized to other surfaces with arbitrary orientation and chemical composition, as well as to other molecular adsorbates. It includes three main steps based on mean-field electronic structure calculations, embedding theories and quantum algorithms. The quantum computing step demonstrated here with the variational quantum eigensolver is amenable to any other reliable quantum algorithm for ground-state energy estimation. Excited-state energies can also be considered in the quantum computing step, if the target reaction involves excited states. 

We are currently working on refining the protocol and extending it to heterogeneous catalysis reactions. We are studying a hydrocarbon cracking reaction on Pt(111). The refinement includes the characterization of multiple interface geometries to compute the potential energy surface at the level of classical computing, exploring different energy estimation and active space reduction schemes at the level of quantum computing. 

  • X. Liu, M. McGrew, R. Di Felice et al., “A Quantum Computational Method for a hydrocarbon cracking reaction on Pt(111)”, manuscript in preparation. 
  • N. Jain and R. Di Felice, “A Quantum Computational Method for Corrosion Inhibition”, Journal of Chemical Theory and Computation 21, 5697-5711 (2025). 

Other applications of quantum computing in chemistry, biology and materials science

Explored computational instances include energy estimation, quantum dynamics and optimization. Hardware includes D-Wave quantum annealers and IBM superconducting devices, accessible on the cloud through the USC IBM quantum innovation center. 

research 5

(a) Left panel: Stick-and-ball models of free-standing (top) and 3-vacancy defect (bottom) graphene configurations. Carbon atoms are represented in white or in red if bordering a defect, missing carbon atoms are indicated by red crosses, while dangling bonds are indicated by black dotted lines. Right panel: Representations of defect-free (top) and 3-vacancy defect (bottom) graphene configurations as an order-3 bipartite graph with two disjointed parts corresponding to the inequivalent carbon atoms in the Bravais lattice. Carbon atoms at the border of the defect are not highlighted here. (b) Energy spectrum of the QUBO formulation. The ground state coincides with the free-standing graphene configuration; the 1st excited state refers to 1-defect graphene configuration; 2nd and 3rd refer to 2-defect graphene configurations, where the lower in energy (2nd) corresponds to the clustered 2-vacancy configuration and the higher (3rd) to the configuration made by two separated 1-vacancies; 4th, 5th, and 6th refer to 3-defect graphene configurations, where, again, the lower in energy correspond to the clustered 3-vacancy configuration. The energy values are reported as differences with respect to the ground state energy. More extended vacancy defects are not considered because of the low probability to be experimentally detected due to their higher instability. For each state, a pictorial representation of the configuration is given. A suitable choice for the QUBO parameters permits DW2Q6 to detect many relevant configurations. From Carnevali et al. (2020). 

 

research 6 Illustration of the principles and purpose of this work. a Top: Quantum versus classical adiabatic annealing processes. With quantum annealing there is the possibility for the system state (red) to tunnel through a changing barrier (black) and arrive at the ground state; for classical annealing, the system must rely on thermal fluctuations (temperature T > 0) to overcome any energy barriers. Bottom: Typical spectrum of instantaneous energy eigenvalues during adiabatic quantum optimization. The ground state energy at s = 0 has a significant gap to the next energy level. The speed at which the optimization can take place depends on the size of the minimum gap. b Using simplified data sets of a small number of sequences derived from actual binding affinity experiments, we use D-Wave, a commercially available quantum annealer, to classify and rank binding affinity preferences of a TF to DNA. From Li et al. (2018) 

 

research 7 Computed ccpvdz GSEE with different algorithms and active spaces, relative to the ground-truth ccpvdz SHCI GSEE. The solid circles represent results obtained from purely classical computational devices, with classical and hybrid classical/quantum algorithms. The hollow circles represent results obtained from a combination of classical and quantum computational devices, with hybrid classical/quantum algorithms. The inset is a zoom on the results obtained with classical algorithms SHCI, DMRG and QMC, showing the small discrepancy that is not resolved on the larger energy scale. From Mehra et al. (2026). 

 

  • D. Mehra, C. Wang, R. Di Felice, “Benchmarking ground-state energy estimation of strongly correlated molecules: the case of the NO single radical”, manuscript under review after revision. 
  • Y. Hang, R. Di Felice, A. Nakano, S. Haas, “Dynamics of Majorana fermions on a quantum computer”, Phys. Rev. B 113, 235145 (2026). 
  • M. Mercado, K. Chen, P. Darekar, A. Nakano, R. Di Felice, S. Haas, “Dynamics of symmetry-protected topological matter on a quantum computer”, Phys Rev. B 110, 075116 (2024). 
  • C. Aiello, D. Awschalom, H. Bernien, T. Bower-Thomas, K. Brown, T. Brun, J. Caram, E. Chitambar, R. Di Felice, K. Edmonds, M. Fox, H. Lewandowski, D. McClure, J. Palsberg, G. Passante, K. Pudenz, C. Richardson, J. Rosenberg, R. Ross, M. Saffman, M. Singh, D. Steuerman, C. Stark, J. Thijssen, N. Vamivakas, J. Whitfield, B. Zwicki, “Achieving a quantum smart workforce”, Quantum Sci. Technol. 6, 030501 (2021). 
  • I. SIloi, V. Carnevali, B. Pokharel, M. Fornari, R. Di Felice*, “Investigating the Chinese Postman Problem on a Quantum Annealer”, Quantum Machine Intelligence 3, 3 (2021).  
  • V. Carnevali, I. Siloi, R. Di Felice, M. Fornari, “Vacancies in graphene: an application of adiabatic quantum optimization”, Phys. Chem. Chem. Phys. 22, 27332-27337 (2020). 
  • R. Li, R. Di Felice, R. Rohs, D. Lidar, “Quantum versus classical machine learning applied to a simplified computational biology problem”, NPJ QI 4, 14 (2018).