The OpenMolcas Web: A Community-Driven Approach to Advancing Computational Chemistry.

The developments of the open-source OpenMolcas chemistry software environment since spring 2020 are described, with a focus on novel functionalities accessible in the stable branch of the package or via interfaces with other packages. These developments span a wide range of topics in computational chemistry and are presented in thematic sections: electronic structure theory, electronic spectroscopy simulations, analytic gradients and molecular structure optimizations, ab initio molecular dynamics, and other new features. This report offers an overview of the chemical phenomena and processes OpenMolcas can address, while showing that OpenMolcas is an attractive platform for state-of-the-art atomistic computer simulations.

Investigation of the pharmacological treatment patterns of Chinese patients with major depressive disorder under real-world settings using multi-channel sequence analysis.

Background: Despite many treatment guidelines available now, the treatment patterns of major depressive disorder (MDD) in China haven't been well-understood due to complexity and diversity. Aim: To describe pharmacological treatment patterns of MDD patients in real-world settings using electronic health records from a major psychiatric hospital in China. Methods: MDD patients (18-65 years, ICD-10: F32.x, F33.x) newly initiated single antidepressant (AD) in 2015 were enrolled, the date of first AD prescription during the study period was defined as index date, and eligible patients were followed up to 1 year. Treatment patterns were revealed and analyzed using multi-channel sequence analysis (MCSA), considering patients' chronological sequences (in days) of AD prescription, cumulative treatment step(s), and polypharmacy usage during the follow-up. Results: This study (n = 5,003) identified four types of MDD treatment patterns. The first type (1-time treatment) represents the largest proportion of patients (73.6%, n = 3,686), followed by the second type (6-month consistent treatment) and third type (long-term, consistent treatment) collectively accounted for 20.6% (n = 1,031) of patients, by contrast the last type (long-term, inconsistent treatment) made up the rest 5.7% (n = 286) of patients while exhibiting the most complicated treatments patterns. The choice of AD was dominated by selective serotonin reuptake inhibitors (SSRIs), while treatment duration spent in polypharmacy spanned at 2.8%, 16.4%, 2.0%, and 36.5% over the four types, respectively. Conclusion: Treatment patterns reflecting real-world pharmacological treatment practices of MDD in China were revealed using MCSA. The observed discrepancies between real-world practice and treatment guidelines provided additional insights in improving the clinical management of MDD.

Automated Active Space Selection with Dipole Moments.

Multireference calculations can provide accurate information of systems with strong correlation, which have increasing importance in the development of new molecules and materials. However, selecting a suitable active space for multireference calculations is nontrivial, and the selection of an unsuitable active space can sometimes lead to results that are not physically meaningful. Active space selection often requires significant human input, and the selection that leads to reasonable results often goes beyond chemical intuition. In this work, we have developed and evaluated two protocols for automated selection of the active space for multireference calculations based on a simple physical observable, the dipole moment, for molecules with nonzero ground-state dipole moments. One protocol is based on the ground-state dipole moment, and the other is based on the excited-state dipole moments. To evaluate the protocols, we constructed a dataset of 1275 active spaces from 25 molecules, each with 51 active space sizes considered, and have mapped out the relationship between the active space, dipole moments, and vertical excitation energies. We have demonstrated that, within this dataset, our protocols allow one to choose among a number of accessible active spaces one that is likely to give reasonable vertical excitation energies, especially for the first three excitations, with no parameters manually decided by the user. We show that, with large active spaces removed from consideration, the accuracy is similar and the time-to-solution can be reduced by more than 10 fold. We also show that the protocols can be applied to potential energy surface scans and determining the spin states of transition metal oxides.

A mechanistic study on the cellular uptake, intracellular trafficking, and antisense gene regulation of bottlebrush polymer-conjugated oligonucleotides.

We have developed a non-cationic transfection vector in the form of bottlebrush polymer-antisense oligonucleotide (ASO) conjugates. Termed pacDNA (polymer-assisted compaction of DNA), these agents show improved biopharmaceutical characteristics and antisense potency in vivo while suppressing non-antisense side effects. Nonetheless, there still is a lack of the mechanistic understanding of the cellular uptake, subcellular trafficking, and gene knockdown with pacDNA. Here, we show that the pacDNA enters human non-small cell lung cancer cells (NCI-H358) predominantly by scavenger receptor-mediated endocytosis and macropinocytosis and trafficks via the endolysosomal pathway within the cell. The pacDNA significantly reduces a target gene expression (KRAS) in the protein level but not in the mRNA level, despite that the transfection of certain free ASOs causes ribonuclease H1 (RNase H)-dependent degradation of KRAS mRNA. In addition, the antisense activity of pacDNA is independent of ASO chemical modification, suggesting that the pacDNA always functions as a steric blocker.

CFD-based pesticide selection for a nozzle used in a six-rotor UAV in hover mode for tea spraying.

BACKGROUND: A key challenge for unmanned aerial vehicle (UAV) spraying sometimes used in tea plantations is the downwash flow structure there stronger than in crops. In addition, the UAV spray is affected by the relationship between the nozzle design and the pesticide. However, there is little current research on this aspect. As a preliminary step this study focuses on the most appropriate pesticide for a designated nozzle in a six-rotor UAV according to the nozzle-pesticide relationship using a three-dimensional computational fluid dynamics model. This model considers the downwash flow structure effect and nozzle spray performance in hover. Nozzle FVP110-02, widely used in six-rotor UAVs, is used as a representative nozzle and bifenthrin and tea saponin water, commonly used in tea plantations, are used as the pesticides. RESULTS: The downwash flow structure of the six-rotor UAV in hover was conveniently controlled by the flight height and rotational speed, thereby causing the turbulence to be more stable. For nozzle FVP110-02, bifenthrin was more appropriate than tea saponin water at the same concentration, whilst bifenthrin and tea saponin water at a concentration of 1:1000 showed the best performance under identical working conditions. CONCLUSION: The numerical model developed here was shown to be effective for investigating the relationship between nozzle and pesticide. Our findings will help to not only improve UAV spraying for tea cultivation but also provide guidelines for pesticide selection in crops. Further work will address the comparison of the rigorous qualification of the numerical simulations with the measurements by the field test. © 2023 Society of Chemical Industry.

Bioaugmentation performances with a powerful strain for nitrogen removal without N2O accumulation.

N2O is regarded as an inevitable intermediate during nitrogen removal, especially for wastewater treatment plants where good operating conditions would be required to mitigate N2O releasing, which generally causes a high treatment cost. In this study, a novel bacterium capable of removing nitrogen without N2O accumulation was isolated and identified as Citrobacter freundii XY-1. The nitrogen removal characteristics, nitrogen removal pathway, bioaugmentation in different reactors as well as microbial diversity were investigated. Results showed that 99.42% of NH+ 4-N and 95% of total organic carbon could be removed within 48 h with the corresponding removal rates being 4.03 mg/(L·h) and 39.42 mg/(L·h), respectively. It was inferred that traditional denitrification and N2O generation do not exist in the pathway of removing nitrogen by XY-1 based on isotope analysis and functional genes detection. Bioaugmentations of XY-1 in both sequencing batch reactor and biological aerated filter significantly promoted the performances of nitrogen removal. The microbial diversity indicated that the relative abundance of strain XY-1 ranged from 45% to 66%, predominating throughout the running period. Overall, XY-1 could become an incredibly important candidate for the upgrading of wastewater treatment plants.

Concordance of the treatment patterns for major depressive disorders between the Canadian Network for Mood and Anxiety Treatments (CANMAT) algorithm and real-world practice in China.

Background: Antidepressant (AD) algorithm is an important tool to support treatment decision-making and improve management of major depressive disorder (MDD). However, little is known about its concordance with real-world practice. This study aimed to assess the concordance between the longitudinal treatment patterns and AD algorithm recommended by a clinical practice guideline in China. Methods: Data were obtained from the electronic medical records of Shanghai Mental Health Center (SMHC), one of the largest mental health institutions in China. We examined the concordance between clinical practice and the Canadian Network for Mood and Anxiety Treatments (CANMAT) algorithm among a cohort composed of 19,955 MDD patients. The longitudinal characteristics of treatment regimen and duration were described to identify the specific inconsistencies. Demographics and health utilizations of the algorithm-concordant and -discordant subgroups with optimized treatment were measured separately. Results: The overall proportion of algorithm-concordant treatment significantly increased from 84.45% to 86.03% during the year of 2015-2017. Among the patients who received recommended first-line drugs with subsequent optimized treatment (n = 2977), the concordance proportion was 27.24%. Mirtazapine and trazodone were the most used drugs for adjunctive strategy. Inadequate or extended duration before optimized treatment are common inconsistency. The median length of follow-up for algorithm-concordant (n = 811) and algorithm-discordant patients (n = 2166) were 153 days (Q1-Q3 = 79-328) and 368 days (Q1-Q3 = 181-577) respectively, and the average number of clinical visits per person-year was 13.07 and 13.08 respectively. Conclusion: Gap existed between clinical practice and AD algorithm. Improved access to evidence-based treatment is required, especially for optimized strategies during outpatient follow-up.

A Long-Circulating Vector for Aptamers Based upon Polyphosphodiester-Backboned Molecular Brushes.

Aptamers face challenges for use outside the ideal conditions in which they are developed. These difficulties are most palpable in vivo due to nuclease activities, rapid clearance, and off-target binding. Herein, we demonstrate that a polyphosphodiester-backboned molecular brush can suppress enzymatic digestion, reduce non-specific cell uptake, enable long blood circulation, and rescue the bioactivity of a conjugated aptamer in vivo. The backbone along with the aptamer is assembled via solid-phase synthesis, followed by installation of poly(ethylene glycol) (PEG) side chains using a two-step process with near-quantitative efficiency. The synthesis allows for precise control over polymer size and architecture. Consisting entirely of building blocks that are generally recognized as safe for therapeutics, this novel molecular brush is expected to provide a highly translatable route for aptamer-based therapeutics.

Targeting oncogenic KRAS with molecular brush-conjugated antisense oligonucleotides.

The mutant form of the guanosine triphosphatase (GTPase) KRAS is a key driver in human tumors but remains a challenging therapeutic target, making KRASMUT cancers a highly unmet clinical need. Here, we report a class of bottlebrush polyethylene glycol (PEG)-conjugated antisense oligonucleotides (ASOs) for potent in vivo KRAS depletion. Owing to their highly branched architecture, these molecular nanoconstructs suppress nearly all side effects associated with DNA-protein interactions and substantially enhance the pharmacological properties of the ASO, such as plasma pharmacokinetics and tumor uptake. Systemic delivery to mice bearing human non-small-cell lung carcinoma xenografts results in a significant reduction in both KRAS levels and tumor growth, and the antitumor performance well exceeds that of current popular ASO paradigms, such as chemically modified oligonucleotides and PEGylation using linear or slightly branched PEG. Importantly, these conjugates relax the requirement on the ASO chemistry, allowing unmodified, natural phosphodiester ASOs to achieve efficacy comparable to that of chemically modified ones. Both the bottlebrush polymer and its ASO conjugates appear to be safe and well tolerated in mice. Together, these data indicate that the molecular brush-ASO conjugate is a promising therapeutic platform for the treatment of KRAS-driven human cancers and warrant further preclinical and clinical development.

Hepatoprotective Effects of Albumin-Encapsulated Nanoparticles of a Curcumin Derivative COP-22 against Lipopolysaccharide/D-Galactosamine-Induced Acute Liver Injury in Mice.

Acute liver injury (ALI) is a severe syndrome and can further develop into acute liver failure (ALF) which can lead to high mortality and cause irreversible liver injuries in the clinic. Liver transplantation is the most common treatment; however, liver donors are lacking, and the progression of ALF is rapid. Nanoparticles can increase the bioavailability and the targeted accumulation of drugs in the liver, so as to significantly improve the therapeutic effect of ALI. Curcumin derivative COP-22 exhibits low cytotoxicity and effective anti-inflammatory activity; however, it has poor water solubility. In this study, COP-22-loaded bovine serum albumin (BSA) nanoparticles (22 NPs) were prepared and characterized. They exhibit effective hepatoprotective effects by inhibiting inflammation, oxidative stress, and apoptosis on Lipopolysaccharide/D-Galactosamine-induced acute liver injury of mice. The anti-inflammatory activity of 22 NPs is related to the regulation of the NF-κB signaling pathways; the antioxidant activity is related to the regulation of the Nrf2 signaling pathways; and the apoptosis activity is related to mitochondrial pathways, involving Bcl-2 family and Caspase-3 protein. These three cellular pathways are interrelated and affected each other. Moreover, 22 NPs could be passively targeted to accumulate in the liver through the retention effect and are more easily absorbed than 22.HCl salt in the liver.

One-year incidence rate of Treatment Resistant Depression (TRD) and treatment characteristics in China.

BACKGROUND: Little is known about the characteristics of Treatment-Resistant Depression (TRD) in China. In previous studies various identification approaches have led to a wide range of results, and it is unclear how Chinese patients compare to those in other studies. METHODS: This is a retrospective cohort study using electronic health records (EHR) from two major psychiatric hospitals in China. Adult major depressive disorder (MDD) patients who initiated pharmaceutical treatment during 2010-2018 were enrolled and follow-up was 1 year. TRD was primarily identified by consensus definition of two failures of adequate (≥4 weeks) regimens. Alternative regimens of 2-weeks and 6-weeks duration, and a data-driven definition were also applied. RESULTS: In the two hospitals, 12,257 (mean age: 40.8y, 63.6% female) and 8314 (mean age: 42.4y, 68.4% female) eligible patients were included. The 1-year incidence rate of TRD was estimated to be 5.2%-7.7% using the primary definition. TRD patients had mean treatment duration of 302.5 days and 285.7 days; had 3.6 and 3.7 treatment steps on average; 94.0% and 72.6% were prescribed polypharmacy regimens, which were all marginally greater than that of non-TRD patients. Alternative definitions resulted in a wide range of incidence estimates (0.5%-20.0%). LIMITATIONS: Medications were assumed to be consumed as prescribed and lack of rating scales from EHRs may limit our TRD identification. CONCLUSIONS: The incidence of TRD among Chinese MDD patients was comparable to other countries under similar settings and more complex treatment characteristics were observed among TRD patients. Alternative TRD definitions revealed the need for better treatment management in practices.

Machine learning dielectric screening for the simulation of excited state properties of molecules and materials.

Accurate and efficient calculations of absorption spectra of molecules and materials are essential for the understanding and rational design of broad classes of systems. Solving the Bethe-Salpeter equation (BSE) for electron-hole pairs usually yields accurate predictions of absorption spectra, but it is computationally expensive, especially if thermal averages of spectra computed for multiple configurations are required. We present a method based on machine learning to evaluate a key quantity entering the definition of absorption spectra: the dielectric screening. We show that our approach yields a model for the screening that is transferable between multiple configurations sampled during first principles molecular dynamics simulations; hence it leads to a substantial improvement in the efficiency of calculations of finite temperature spectra. We obtained computational gains of one to two orders of magnitude for systems with 50 to 500 atoms, including liquids, solids, nanostructures, and solid/liquid interfaces. Importantly, the models of dielectric screening derived here may be used not only in the solution of the BSE but also in developing functionals for time-dependent density functional theory (TDDFT) calculations of homogeneous and heterogeneous systems. Overall, our work provides a strategy to combine machine learning with electronic structure calculations to accelerate first principles simulations of excited-state properties.

The Dissolution Rate of CaCO3 in the Ocean.

The dissolution of CaCO3 minerals in the ocean is a fundamental part of the marine alkalinity and carbon cycles. While there have been decades of work aimed at deriving the relationship between dissolution rate and mineral saturation state (a so-called rate law), no real consensus has been reached. There are disagreements between laboratory- and field-based studies and differences in rates for inorganic and biogenic materials. Rates based on measurements on suspended particles do not always agree with rates inferred from measurements made near the sediment-water interface of the actual ocean. By contrast, the freshwater dissolution rate of calcite has been well described by bulk rate measurements from a number of different laboratories, fit by basic kinetic theory, and well studied by atomic force microscopy and vertical scanning interferometry to document the processes at the atomic scale. In this review, we try to better unify our understanding of carbonate dissolution in the ocean via a relatively new, highly sensitive method we have developed combined with a theoretical framework guided by the success of the freshwater studies. We show that empirical curve fits of seawater data as a function of saturation state do not agree, largely because the curvature is itself a function of the thermodynamics. Instead, we show that models that consider both surface energetic theory and the complicated speciation of seawater and calcite surfaces in seawater are able to explain most of the most recent data.This new framework can also explain features of the historical data that have not been previously explained. The existence of a kink in the relationship between rate and saturation state, reflecting a change in dissolution mechanism, may be playing an important role in accelerating CaCO3 dissolution in key sedimentary environments.

OpenMolcas: From Source Code to Insight.

In this Article we describe the OpenMolcas environment and invite the computational chemistry community to collaborate. The open-source project already includes a large number of new developments realized during the transition from the commercial MOLCAS product to the open-source platform. The paper initially describes the technical details of the new software development platform. This is followed by brief presentations of many new methods, implementations, and features of the OpenMolcas program suite. These developments include novel wave function methods such as stochastic complete active space self-consistent field, density matrix renormalization group (DMRG) methods, and hybrid multiconfigurational wave function and density functional theory models. Some of these implementations include an array of additional options and functionalities. The paper proceeds and describes developments related to explorations of potential energy surfaces. Here we present methods for the optimization of conical intersections, the simulation of adiabatic and nonadiabatic molecular dynamics, and interfaces to tools for semiclassical and quantum mechanical nuclear dynamics. Furthermore, the Article describes features unique to simulations of spectroscopic and magnetic phenomena such as the exact semiclassical description of the interaction between light and matter, various X-ray processes, magnetic circular dichroism, and properties. Finally, the paper describes a number of built-in and add-on features to support the OpenMolcas platform with postcalculation analysis and visualization, a multiscale simulation option using frozen-density embedding theory, and new electronic and muonic basis sets.

Nature of the 11Bu and 21Ag Excited States of Butadiene and the Goldilocks Principle of Basis Set Diffuseness.

Butadiene, being the simplest conjugated organic molecule, has been studied extensively by experiments and various high-level theoretical methods. Previous studies conclude that the complete active space (CAS) self-consistent field (SCF) method was unable to obtain the correct 11Bu and 21Ag state ordering and that it introduces artificial valence-Rydberg mixing into the 11Bu state because of the lack of external correlation. Basis sets and initial guesses specifically constructed for this problem were able to improve the vertical excitation energy of the 11Bu state but did not resolve the controversy of the nature of the 11Bu state. In the present work, we demonstrate that, using standard intermediately diffuse basis sets such as jul-cc-pVTZ and ma-TZVP, CASSCF is able to obtain the 11Bu and 21Ag states of trans-butadiene with much improved vertical excitation energy and reasonable wave function characteristics, and it provides a good reference wave function (capable of yielding quantitative excitation energies for excited states) for three methods that treat electron correlation in different ways, namely, multiconfiguration pair-density functional theory (MC-PDFT), CAS second-order perturbation theory (PT2), and multistate (MS) CASPT2. We demonstrate that a combined analysis of the orbital second moments, state second moments, and MC-PDFT energy components is a very useful approach in determining excited-state characteristics and assigning states, and we show that basis sets without diffuse functions or with very diffuse basis functions are not stable or accurate in predicting the excited states of butadiene. We show that the 21Ag state is valence-like and has an atypical double/single excitation character and the 11Bu state has a certain degree of Rydberg character that is not artificial, settling the decades of controversy of the characters of these states.

State-Interaction Pair-Density Functional Theory Can Accurately Describe a Spiro Mixed Valence Compound.

Mixed-valence compounds with strong couplings between electronic states constitute one of the most challenging types of multireference systems for electronic structure theory. Previous work on a model mixed-valence compound, the 2,2',6,6'-tetrahydro-4 H,4' H-5,5'-spirobi[cyclopenta[ c]pyrrole] cation, showed that multireference perturbation theory (MRPT) can give a physical energy surface for the mixed-valence compound only by going to the third order or by using a scheme involving averaging orbital energies in a way specific to mixed-valence systems. In this study, we show that second-order MRPT methods (CASPT2, MS-CASPT2, and XMS-CASPT2) can give good results by calculating the Fock operator for the zeroth-order Hamiltonian using the state-averaged density matrix. We also show that state-interaction pair-density functional theory (SI-PDFT) is free from the unphysical behavior of previously tested second-order MRPT methods for this prototype mixed-valence compound near the avoided crossing. This is very encouraging because of the much lower cost in applying SI-PDFT to large or complex systems.

Assessment of MC-PDFT Excitation Energies for a Set of QM/MM Models of Rhodopsins.

A methodology for the automatic production of quantum mechanical/molecular mechanical (QM/MM) models of retinal-binding rhodopsin proteins and subsequent prediction of their spectroscopic properties has been proposed recently by some of the authors. The technology employed for the evaluation of the excitation energies is called Automatic Rhodopsin Modeling (ARM), and it involves the use of the complete active space self-consistent field (CASSCF) method followed by a multiconfiguration second-order perturbation theory (in particular, CASPT2) calculation of external correlation energies. Although it was shown that ARM is capable of successfully reproducing and predicting spectroscopic property trends in chromophore-embedding protein sets, practical applications of such technology are limited by the high computational costs of the multiconfiguration perturbation theory calculations. In the present work we benchmark the more affordable multiconfiguration pair-density functional theory (MC-PDFT) method whose accuracy has been recently validated for retinal chromophores in the gas phase, indicating that MC-PDFT could potentially be used to analyze large (e.g., few hundreds) sets of rhodopsin proteins. Here, we test this theory for a set of rhodopsin QM/MM models whose experimental absorption maxima (λ a max) have been measured. The results indicate that MC-PDFT may be employed to calculate λ a max values for this important class of photoresponsive proteins.

The kinetics of siderophore-mediated olivine dissolution.

Silicate minerals represent an important reservoir of nutrients at Earth's surface and a source of alkalinity that modulates long-term geochemical cycles. Due to the slow kinetics of primary silicate mineral dissolution and the potential for nutrient immobilization by secondary mineral precipitation, the bioavailability of many silicate-bound nutrients may be limited by the ability of micro-organisms to actively scavenge these nutrients via redox alteration and/or organic ligand production. In this study, we use targeted laboratory experiments with olivine and the siderophore deferoxamine B to explore how microbial ligands affect nutrient (Fe) release and the overall rate of mineral dissolution. Our results show that olivine dissolution rates are accelerated in the presence of micromolar concentrations of deferoxamine B. Based on the non-linear decrease in rates with time and formation of a Fe3+ -ligand complex, we attribute this acceleration in dissolution rates to the removal of an oxidized surface coating that forms during the dissolution of olivine at circum-neutral pH in the presence of O2 and the absence of organic ligands. While increases in dissolution rates are observed with micromolar concentrations of siderophores, it remains unclear whether such conditions could be realized in natural environments due to the strong physiological control on microbial siderophore production. So, to contextualize our experimental results, we also developed a feedback model, which considers how microbial physiology and ligand-promoted mineral dissolution kinetics interact to control the extent of biotic enhancement of dissolution rates expected for different environments. The model predicts that physiological feedbacks severely limit the extent to which dissolution rates may be enhanced by microbial activity, though the rate of physical transport modulates this limitation.

Changes in the epidemic of pulmonary tuberculosis in Shanghai from 1992 to 2016.

OBJECTIVE: To estimate whether WHO's End TB Strategy targets can be achieved by analysing the incidence trend of pulmonary tuberculosis in Shanghai during 1992-2016. METHOD: The age-adjusted annual incidence of tuberculosis (TB) was calculated based on data from the national TB registration system. Linear regression was applied to analyse the trend of the epidemic, together with the statistical indicator of annual percent change (APC). RESULTS: The overall age-standardised rate decreased from 34.8/100 000 in 1992 to 21.2/100 000 in 2016, or by 2.15% (t = -13.258, P < 0.05) annually. After rapidly declining between 1999 and 2003 (-5.4% p.a.), the epidemic remained at a stable level with a lower annual declining rate (-1.1% p.a). In 2035, the estimated incidence will be 17.2/100 000 based on the APC in 2004-2016. There were two peaks in average incidence of the total population, 30.9/100 000 in the 20-24 age group and 66.4/100 000 in the 70-74 age group. Overall, the ratio of new to retreated cases continually rose and eventually reached 8.36:1 in 2016. The constituent ratio of smear-positive cases ranged from 35.9% to 47.8% without rising or decreasing trend (P = 0.065). CONCLUSION: The epidemic of TB in Shanghai has steadily declined during last two decades. A new strategy should be developed to rapidly reduce the incidence rate to achieve the WHO Goals in 2035.

Extended Hamiltonian molecular dynamics: semiclassical trajectories with improved maintenance of zero point energy.

It is well known that classical trajectories, even if they are initiated with zero point energy (ZPE) in each mode (trajectories initiated this way are commonly called quasiclassical trajectories), do not maintain ZPE in the final states. The energy of high-frequency modes will typically leak into low-frequency modes or relative translation of subsystems during the time evolution. This can lead to severe problems such as unphysical dissociation of a molecule, production of energetically disallowed reaction products, and unphysical product energy distributions. Here a new molecular dynamics method called extended Hamiltonian molecular dynamics (EHMD) is developed to improve the ZPE problem in classical molecular dynamics. In EHMD, two images of a trajectory are connected by one or more springs. The EHMD method is tested with the Henon-Heiles Hamiltonian in reduced and real units and with a Hamiltonian with quartic anharmonicity in real units, and the method is found to improve zero-point maintenance as intended.