The 2007 fragment molecular orbital (FMO) study of ubiquitin in an explicit water shell (Komeiji, Ishida, Fedorov, and Kitaura) established, at the HF/MP2 ab initio level, that solvation induces a residue-resolved charge transfer of order one electron between protein and solvent, a large dipole enhancement, and an internal-destabilization versus interaction-stabilization energy balance — but only as single-point calculations on a small set of energy-minimized configurations, with Mulliken charges and an excised 12 Å solvent sphere. Single-point FMO on minimized snapshots does not incorporate temperature and configurational fluctuation, and full ab initio FMO-MD remains future work (Komeiji et al. 2012; Okiyama et al. 2020). The open problem is to quantify the dynamical modulation of these observables under thermal sampling (FMO-MD), removing droplet boundary artifacts via periodic boundary conditions, controlling population-analysis and basis-set uncertainty, delineating explicit versus implicit solvent, and connecting the analysis to ab initio solvation free energies. FMO-MD exists but only for small solutes; periodic-boundary FMO was implemented at the density-functional-tight-binding level (2021–2022), including protein crystals and free-energy decomposition; implicit-solvent FMO-PBSA comparisons and statistically averaged MD-FMO ligand-binding analyses are now routine. What remains open is an all-electron fragment-based ab initio dynamical treatment of a fully solvated protein with converged bulk-solvent boundary conditions, statistically converged electronic observables, quantified population-scheme and basis-set robustness, and the associated ab initio solvation thermodynamics, with the source ubiquitin system as benchmark. As of 2026 this core remains open.

Background

The fragment molecular orbital (FMO) method is a quantum-chemical fragmentation scheme that extends ab initio electronic-structure theory — Hartree–Fock (HF), second-order Møller–Plesset perturbation theory (MP2), or density-functional theory — to systems of thousands of atoms (Fedorov, Nagata, and Kitaura, PCCP 2012). The molecule is divided into fragments (for proteins, typically one amino-acid residue or one solvent cluster per fragment); each fragment's wave function is computed ab initio in the electrostatic field of all other fragments, and dimer corrections recover interfragment interactions including many-body polarization. FMO yields, at tractable cost, per-fragment observables — Mulliken populations (electron counts assigned from basis-function overlap, a scheme known to depend on basis set and partitioning convention) and fragment dipole moments — as well as a pair interaction energy (PIE) analysis, decomposable (PIEDA) into electrostatic, exchange-repulsion, charge-transfer, and dispersion components.

In 2007 Komeiji, Ishida, Fedorov, and Kitaura used FMO at the HF and MP2 levels with the 6-31G* basis set to compare ubiquitin (a 76-residue model protein) in vacuo with the same protein immersed in an explicit water shell up to 12 Å thick. The configurations were energy-minimized snapshots drawn from a classical molecular dynamics (MD) trajectory of the solvated protein, and all electronic calculations were single-point (fixed nuclei). The study reported that solvation causes a net transfer of electronic charge of order one electron between protein and solvent (a net change of the protein's Mulliken charge of order one atomic unit of charge), with positively charged surface residues (Arg, Lys) drawing electrons from the water and negatively charged ones (Asp, Glu) releasing electrons to it; that the protein dipole, obtained as the vector sum of fragment dipoles, is markedly enhanced upon solvation; that the protein's internal energy is destabilized by solvation, more than compensated by the protein–solvent interaction; and that correlation (MP2) contributions to the energetics saturate within the first solvation shell whereas the electrostatic (HF) contributions converge more slowly with shell thickness — in the same authors' later account of this study, the net charge converged at a 4 Å shell, the internal energy at 8 Å, and the protein–solvent interaction showed marginal convergence around 8 Å.

This picture was static and methodologically fragile: only a few minimized configurations; an excised solvent sphere whose outer surface introduces unphysical boundary artifacts and whose thickness was swept on a single configuration; Mulliken charges, which are basis-set and scheme dependent; fragment-summed dipoles that neglect cross-fragment polarization; and MP2 values uncorrected for basis-set superposition error (BSSE, the artificial stabilization obtained by using neighboring fragments' basis functions). The unresolved problem is to turn this snapshot description into a dynamical, statistically converged, boundary-artifact-free, and quantitatively controlled ab initio description connected to solvation thermodynamics.

Neighboring approaches define the landscape. In ab initio (Born–Oppenheimer) molecular dynamics, nuclear forces are computed from quantum chemistry on the fly, so electronic reorganization feeds back into nuclear motion; it has been applied to solvated proteins at density-functional levels but without residue-resolved fragment decomposition or MP2-class correlation. Implicit-solvent models (Poisson–Boltzmann and polarizable continuum methods) replace explicit water by a dielectric reaction field and are far cheaper, but cannot describe residue-specific charge transfer. Periodic boundary conditions (PBC), in which the simulation cell is replicated throughout space, remove the artificial surface of an excised droplet and provide a genuine bulk solvent environment. FMO-MD — FMO-computed forces driving molecular dynamics — was developed by the same group and demonstrated on small solutes, and periodic FMO formulations at the semiempirical density-functional tight-binding (DFTB) level appeared in 2021–2022. What has not been achieved is the combination at the all-electron ab initio level for a whole solvated protein.

Problem Statement

Determine, at all-electron ab initio fragment-based accuracy (the HF/MP2 class of the source study or better), the electronic structure of a solvated protein — the source ubiquitin system serving as the benchmark — as a dynamical, statistically converged object, replacing the 2007 single-point, excised-droplet description of ubiquitin. Concretely, the problem asks to:

(i) carry out fragment-based ab initio molecular dynamics (FMO-MD) of the benchmark solvated protein with explicit solvent in a converged bulk-solvent environment — periodic boundary conditions or a demonstrated equivalent that eliminates the outer-surface artifact of excised shells — and quantify how thermal and conformational fluctuations, and the feedback between electronic reorganization and nuclear motion, modulate the observables established in the source study: residue-resolved and net charge transfer between protein and solvent, the protein dipole moment and its fluctuations, and the partition between internal protein destabilization and protein–solvent interaction;

(ii) establish the methodological robustness of these observables, replacing sole reliance on Mulliken populations and summed fragment dipoles by a systematic assessment of population-scheme and basis-set dependence (including density-based alternatives), so that the absolute charge and dipole values are quantitative benchmarks rather than scheme-dependent conventions;

(iii) connect the electronic-structure observables to ab initio solvation thermodynamics of the solvated protein — solvation free energies — computed in a statistically converged ab initio fragment framework, with the relative merits of explicit and implicit solvent treatments delineated for these observables. Protonation-state equilibria and the energetics of biologically relevant processes such as ligand binding or enzymatic steps are not part of the completion requirement and may serve only as optional illustrations.

These deliverables jointly serve one objective — a converged ab initio description of how a protein's electronic structure responds to being in bulk water — and an answer is complete when the charge-transfer, polarization, and energetic observables, including the ab initio solvation free energies and the explicit-versus-implicit solvent comparison, are produced from fragment-based ab initio dynamics with electronic–nuclear feedback and known with quantified statistical and methodological uncertainty, and the qualitative picture reported in the 2007 source study for ubiquitin (direction and residue-resolved pattern of protein–solvent charge transfer, dipole enhancement upon solvation, and internal destabilization more than compensated by protein–solvent interaction) is reproduced in the minimized-configuration limit under a specified population-analysis scheme. 'Adequate statistics' and 'converged bulk solvent' are to be judged by the field's conventions (for example, block-averaged convergence of the reported observables and box-size or shell-size saturation), not by arbitrary numerical thresholds.

References

  1. Yuto Komeiji, Toyokazu Ishida, Dmitri G. Fedorov, Kazuo Kitaura (2007). Change in a protein's electronic structure induced by an explicit solvent: An ab initio fragment molecular orbital study of ubiquitin. Journal of Computational Chemistry 28, 1750-1762. DOI: 10.1002/jcc.20686. https://doi.org/10.1002/jcc.20686
  2. Dmitri G. Fedorov, Takeshi Nagata, Kazuo Kitaura (2012). Exploring chemistry with the fragment molecular orbital method. Physical Chemistry Chemical Physics 14, 7562. DOI: 10.1039/c2cp23784a. https://doi.org/10.1039/c2cp23784a
  3. Yuto Komeiji, Takeshi Ishikawa, Yuji Mochizuki, Hiroshi Yamataka, Tatsuya Nakano (2009). Fragment Molecular Orbital method-based Molecular Dynamics (FMO-MD) as a simulator for chemical reactions in explicit solvation. Journal of Computational Chemistry 30, 40-50. DOI: 10.1002/jcc.21025. https://doi.org/10.1002/jcc.21025
  4. Makoto Sato, Hiroshi Yamataka, Yuto Komeiji, Yuji Mochizuki (2012). FMO-MD Simulations on the Hydration of Formaldehyde in Water Solution with Constraint Dynamics. Chemistry - A European Journal 18, 9714-9721. DOI: 10.1002/chem.201200874. https://doi.org/10.1002/chem.201200874
  5. Kurt R. Brorsen, Noriyuki Minezawa, Feng Xu, Theresa L. Windus, Mark S. Gordon (2012). Fragment Molecular Orbital Molecular Dynamics with the Fully Analytic Energy Gradient. Journal of Chemical Theory and Computation 8, 5008-5012. DOI: 10.1021/ct3007869. https://doi.org/10.1021/ct3007869
  6. Victor M. Anisimov, Vladislav Bugaenko, Claudio N. Cavasotto (2009). Quantum Mechanical Dynamics of Charge Transfer in Ubiquitin in Aqueous Solution. ChemPhysChem 10, 3194-3196. DOI: 10.1002/cphc.200900535. https://doi.org/10.1002/cphc.200900535
  7. Ivan S. Ufimtsev, Nathan Luehr, Todd J. Martínez (2011). Charge Transfer and Polarization in Solvated Proteins from Ab Initio Molecular Dynamics. The Journal of Physical Chemistry Letters 2, 1789-1793. DOI: 10.1021/jz200697c. https://doi.org/10.1021/jz200697c
  8. Yoshio Nishimoto, Dmitri G. Fedorov (2021). The fragment molecular orbital method combined with density-functional tight-binding and periodic boundary conditions. The Journal of Chemical Physics 154, 111102. DOI: 10.1063/5.0039520. https://doi.org/10.1063/5.0039520
  9. Taiji Nakamura, Tomoko Yokaichiya, Dmitri G. Fedorov (2021). Quantum-Mechanical Structure Optimization of Protein Crystals and Analysis of Interactions in Periodic Systems. The Journal of Physical Chemistry Letters 12, 8757-8762. DOI: 10.1021/acs.jpclett.1c02510. https://doi.org/10.1021/acs.jpclett.1c02510
  10. Dmitri G. Fedorov (2022). Polarization energies in the fragment molecular orbital method. Journal of Computational Chemistry 43, 1094-1103. DOI: 10.1002/jcc.26869. https://doi.org/10.1002/jcc.26869
  11. Dmitri G. Fedorov, Taiji Nakamura (2022). Free Energy Decomposition Analysis Based on the Fragment Molecular Orbital Method. The Journal of Physical Chemistry Letters 13, 1596-1601. DOI: 10.1021/acs.jpclett.2c00040. https://doi.org/10.1021/acs.jpclett.2c00040
  12. Dmitri G. Fedorov, Katarzyna J. Zator, Julia Contreras-García, Seiji Mori (2026). Noncovalent Interactions in Solvated Proteins and Protein Crystals Studied with the Fragment Molecular Orbital Method. Journal of Chemical Information and Modeling 66, 1715-1726. DOI: 10.1021/acs.jcim.5c02954. https://doi.org/10.1021/acs.jcim.5c02954
  13. Yoshio Okiyama, Chiduru Watanabe, Kaori Fukuzawa, Yuji Mochizuki, Tatsuya Nakano, Shigenori Tanaka (2019). Fragment Molecular Orbital Calculations with Implicit Solvent Based on the Poisson-Boltzmann Equation: II. Protein and Its Ligand-Binding System Studies. The Journal of Physical Chemistry B 123, 957-973. DOI: 10.1021/acs.jpcb.8b09326. https://doi.org/10.1021/acs.jpcb.8b09326
  14. Yuto Komeij, Yoshio Okiyama, Yuji Mochizuki, Kaori Fukuzawa (2017). Explicit solvation of a single-stranded DNA, a binding protein, and their complex: a suitable protocol for fragment molecular orbital calculation. Chem-Bio Informatics Journal 17, 72-84. DOI: 10.1273/cbij.17.72. https://doi.org/10.1273/cbij.17.72
  15. Ryohei Yoshine, Yoshinori Hirano, Hideo Doi, Shun Kitahara, Kazuki Akisawa, Koji Okuwaki, Eiji Yamamoto, Kenji Yasuoka, Yuji Mochizuki (2026). Residue-resolved dynamically averaged interaction analysis of direct factor Xa inhibitors by MD-FMO combination calculations. Journal of Computer-Aided Molecular Design 40(1), article 132. DOI: 10.1007/s10822-026-00846-x. https://doi.org/10.1007/s10822-026-00846-x
  16. Yuto Komeiji, Shu Ohno, Shuhei Miyakawa, Kaori Fukuzawa (2026). Fragment Molecular Orbital-Based Molecular Dynamics (FMO-MD): Current Status and Challenges. In: Methods in Molecular Biology (Quantum Mechanics Applications in Drug Design), Springer, pp. 301-309. DOI: 10.1007/978-1-0716-5420-0_17. https://doi.org/10.1007/978-1-0716-5420-0_17
  17. Paige E. Bowling, Dustin R. Broderick, John M. Herbert (2025). Convergent Protocols for Computing Protein-Ligand Interaction Energies Using Fragment-Based Quantum Chemistry. Journal of Chemical Theory and Computation 21, 951-966. DOI: 10.1021/acs.jctc.4c01429. https://doi.org/10.1021/acs.jctc.4c01429
  18. Yuto Komeiji, Yuji Mochizuki, Tatsuya Nakano, Hirotoshi Mori (2012). Recent Advances in Fragment Molecular Orbital-Based Molecular Dynamics (FMO-MD) Simulations. In: Molecular Dynamics – Theoretical Developments and Applications in Nanotechnology and Energy, IntechOpen (open access). https://api.intechopen.com/chapter/pdf-download/34975.pdf
  19. Yoshio Okiyama, Kaori Fukuzawa, Yuto Komeiji, Shigenori Tanaka (2020). Taking Water into Account with the Fragment Molecular Orbital Method. In: Quantum Mechanics in Drug Discovery, Springer, pp. 105-122. Accepted manuscript: https://da.lib.kobe-u.ac.jp/da/kernel/90009514/90009514.pdf