Physics-Informed Active Learning for Accelerating Quantum Chemical Simulations.

Quantum chemical simulations can be greatly accelerated by constructing machine learning potentials, which is often done using active learning (AL). The usefulness of the constructed potentials is often limited by the high effort required and their insufficient robustness in the simulations. Here, we introduce the end-to-end AL for constructing robust data-efficient potentials with affordable investment of time and resources and minimum human interference. Our AL protocol is based on the physics-informed sampling of training points, automatic selection of initial data, uncertainty quantification, and convergence monitoring. The versatility of this protocol is shown in our implementation of quasi-classical molecular dynamics for simulating vibrational spectra, conformer search of a key biochemical molecule, and time-resolved mechanism of the Diels-Alder reaction. These investigations took us days instead of weeks of pure quantum chemical calculations on a high-performance computing cluster.

Emerging diabetes therapies: Regenerating pancreatic ß cells.

The incidence of diabetes mellitus (DM) is steadily increasing annually, with 537 million diabetic patients as of 2021. Restoring diminished ß cell mass or impaired islet function is crucial in treating DM, particularly type 1 diabetes mellitus (T1DM). However, the regenerative capacity of islet ß cells, which primarily produce insulin, is severely limited, and natural regeneration is only observed in young rodents or children. Hence, there is an urgent need to develop advanced therapeutic approaches that can regenerate endogenous ß cells or replace them with stem cell (SC)-derived or engineered ß-like cells. Current strategies for treating insulin-dependent DM mainly include promoting the self-replication of endogenous ß cells, inducing SC differentiation, reprogramming non-ß cells into ß-like cells, and generating pancreatic-like organoids through cell-based intervention. In this Review, we discuss the current state of the art in these approaches, describe associated challenges, propose potential solutions, and highlight ongoing efforts to optimize ß cell or islet transplantation and related clinical trials. These effective cell-based therapies will generate a sustainable source of functional ß cells for transplantation and lay strong foundations for future curative treatments for DM.

TNKS1BP1 mediates AECII senescence and radiation induced lung injury through suppressing EEF2 degradation.

BACKGROUND: Although recent studies provide mechanistic understanding to the pathogenesis of radiation induced lung injury (RILI), rare therapeutics show definitive promise for treating this disease. Type II alveolar epithelial cells (AECII) injury in various manner results in an inflammation response to initiate RILI. RESULTS: Here, we reported that radiation (IR) up-regulated the TNKS1BP1, causing progressive accumulation of the cellular senescence by up-regulating EEF2 in AECII and lung tissue of RILI mice. Senescent AECII induced Senescence-Associated Secretory Phenotype (SASP), consequently activating fibroblasts and macrophages to promote RILI development. In response to IR, elevated TNKS1BP1 interacted with and decreased CNOT4 to suppress EEF2 degradation. Ectopic expression of EEF2 accelerated AECII senescence. Using a model system of TNKS1BP1 knockout (KO) mice, we demonstrated that TNKS1BP1 KO prevents IR-induced lung tissue senescence and RILI. CONCLUSIONS: Notably, this study suggested that a regulatory mechanism of the TNKS1BP1/CNOT4/EEF2 axis in AECII senescence may be a potential strategy for RILI.

Low-dose ionizing radiation-induced RET/PTC1 rearrangement via the non-homologous end joining pathway to drive thyroid cancer.

Thyroid cancer incidence increases worldwide annually, primarily due to factors such as ionizing radiation (IR), iodine intake, and genetics. Papillary carcinoma of the thyroid (PTC) accounts for about 80% of thyroid cancer cases. RET/PTC1 (coiled-coil domain containing 6 [CCDC6]-rearranged during transfection) rearrangement is a distinctive feature in over 70% of thyroid cancers who exposed to low doses of IR in Chernobyl and Hiroshima‒Nagasaki atomic bombings. This study aims to elucidate mechanism between RET/PTC1 rearrangement and IR in PTC. N-thy-ori-3-1 cells were subjected to varying doses of IR (2/1/0.5/0.2/0.1/0.05 Gy) of IR at different days, and result showed low-dose IR-induced RET/PTC1 rearrangement in a dose-dependent manner. RET/PTC1 has been observed to promote PTC both in vivo and in vitro. To delineate the role of different DNA repair pathways, SCR7, RI-1, and Olaparib were employed to inhibit non-homologous end joining (NHEJ), homologous recombination (HR), and microhomology-mediated end joining (MMEJ), respectively. Notably, inhibiting NHEJ enhanced HR repair efficiency and reduced IR-induced RET/PTC1 rearrangement. Conversely, inhibiting HR increased NHEJ repair efficiency and subsequent RET/PTC1 rearrangement. The MMEJ did not show a markable role in this progress. Additionally, inhibiting DNA-dependent protein kinase catalytic subunit (DNA-PKcs) decreased the efficiency of NHEJ and thus reduced IR-induced RET/PTC1 rearrangement. To conclude, the data suggest that NHEJ, rather than HR or MMEJ, is the critical cause of IR-induced RET/PTC1 rearrangement. Targeting DNA-PKcs to inhibit the NHEJ has emerged as a promising therapeutic strategy for addressing IR-induced RET/PTC1 rearrangement in PTC.

Sensor Network Attack Synthesis against Fault Diagnosis of Discrete Event Systems.

This paper investigates the problem of synthesizing network attacks against fault diagnosis in the context of discrete event systems (DESs). It is assumed that the sensor observations sent to the operator that monitors a system are tampered with by an active attacker. We first formulate the process of online fault diagnosis under attack. Then, from the attack viewpoint, we define a sensor network attacker as successful if it can degrade the fault diagnosis in the case of maintaining itself as undiscovered by the operator. To verify such an attacker, an information structure called a joint diagnoser (JD) is proposed, which describes all possible attacks in a given attack scenario. Based on the refined JD, i.e., stealthy joint diagnoser (SJD), we present an algorithmic procedure for synthesizing a successful attacker if it exists.

Scalable parameterized quantum circuits classifier.

As a generalized quantum machine learning model, parameterized quantum circuits (PQC) have been found to perform poorly in terms of classification accuracy and model scalability for multi-category classification tasks. To address this issue, we propose a scalable parameterized quantum circuits classifier (SPQCC), which performs per-channel PQC and combines the measurements as the output of the trainable parameters of the classifier. By minimizing the cross-entropy loss through optimizing the trainable parameters of PQC, SPQCC leads to a fast convergence of the classifier. The parallel execution of identical PQCs on different quantum machines with the same structure and scale reduces the complexity of classifier design. Classification simulations performed on the MNIST Dataset show that the accuracy of our proposed classifier far exceeds that of other quantum classification algorithms, achieving the state-of-the-art simulation result and surpassing/reaching classical classifiers with a considerable number of trainable parameters. Our classifier demonstrates excellent scalability and classification performance.

SimPLE, a visuotactile method learned in simulation to precisely pick, localize, regrasp, and place objects.

Existing robotic systems have a tension between generality and precision. Deployed solutions for robotic manipulation tend to fall into the paradigm of one robot solving a single task, lacking "precise generalization," or the ability to solve many tasks without compromising on precision. This paper explores solutions for precise and general pick and place. In precise pick and place, or kitting, the robot transforms an unstructured arrangement of objects into an organized arrangement, which can facilitate further manipulation. We propose SimPLE (Simulation to Pick Localize and placE) as a solution to precise pick and place. SimPLE learns to pick, regrasp, and place objects given the object's computer-aided design model and no prior experience. We developed three main components: task-aware grasping, visuotactile perception, and regrasp planning. Task-aware grasping computes affordances of grasps that are stable, observable, and favorable to placing. The visuotactile perception model relies on matching real observations against a set of simulated ones through supervised learning to estimate a distribution of likely object poses. Last, we computed a multistep pick-and-place plan by solving a shortest-path problem on a graph of hand-to-hand regrasps. On a dual-arm robot equipped with visuotactile sensing, SimPLE demonstrated pick and place of 15 diverse objects. The objects spanned a wide range of shapes, and SimPLE achieved successful placements into structured arrangements with 1-mm clearance more than 90% of the time for six objects and more than 80% of the time for 11 objects.

MLatom Software Ecosystem for Surface Hopping Dynamics in Python with Quantum Mechanical and Machine Learning Methods.

We present an open-source MLatom@XACS software ecosystem for on-the-fly surface hopping nonadiabatic dynamics based on the Landau-Zener-Belyaev-Lebedev algorithm. The dynamics can be performed via Python API with a wide range of quantum mechanical (QM) and machine learning (ML) methods, including ab initio QM (CASSCF and ADC(2)), semiempirical QM methods (e.g., AM1, PM3, OMx, and ODMx), and many types of ML potentials (e.g., KREG, ANI, and MACE). Combinations of QM and ML methods can also be used. While the user can build their own combinations, we provide AIQM1, which is based on Δ-learning and can be used out-of-the-box. We showcase how AIQM1 reproduces the isomerization quantum yield of trans-azobenzene at a low cost. We provide example scripts that, in dozens of lines, enable the user to obtain the final population plots by simply providing the initial geometry of a molecule. Thus, those scripts perform geometry optimization, normal mode calculations, initial condition sampling, parallel trajectories propagation, population analysis, and final result plotting. Given the capabilities of MLatom to be used for training different ML models, this ecosystem can be seamlessly integrated into the protocols building ML models for nonadiabatic dynamics. In the future, a deeper and more efficient integration of MLatom with Newton-X will enable a vast range of functionalities for surface hopping dynamics, such as fewest-switches surface hopping, to facilitate similar workflows via the Python API.

Integrated profiling identifies DXS253E as a potential prognostic marker in colorectal cancer.

BACKGROUND: Increasing evidence suggests that DXS253E is critical for cancer development and progression, but the function and potential mechanism of DXS253E in colorectal cancer (CRC) remain largely unknown. In this study, we evaluated the clinical significance and explored the underlying mechanism of DXS253E in CRC. METHODS: DXS253E expression in cancer tissues was investigated using the Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) databases. The Kaplan-Meier plot was used to assess the prognosis of DXS253E. The cBioPortal, MethSurv, and Tumor Immune Estimation Resource (TIMER) databases were employed to analyze the mutation profile, methylation, and immune infiltration associated with DXS253E. The biological functions of DXS253E in CRC cells were determined by CCK-8 assay, plate cloning assay, Transwell assay, flow cytometry, lactate assay, western blot, and qRT-PCR. RESULTS: DXS253E was upregulated in CRC tissues and high DXS253E expression levels were correlated with poor survival in CRC patients. Our bioinformatics analyses showed that high DXS253E gene methylation levels were associated with the favorable prognosis of CRC patients. Furthermore, DXS253E levels were linked to the expression levels of several immunomodulatory genes and an abundance of immune cells. Mechanistically, the overexpression of DXS253E enhanced proliferation, migration, invasion, and the aerobic glycolysis of CRC cells through the AKT/mTOR pathway. CONCLUSIONS: We demonstrated that DXS253E functions as a potential role in CRC progression and may serve as an indicator of outcomes and a therapeutic target for regulating the AKT/mTOR pathway in CRC.

DNA-PKcs/AKT1 inhibits epithelial-mesenchymal transition during radiation-induced pulmonary fibrosis by inducing ubiquitination and degradation of Twist1.

INTRODUCTION: Radiation-induced pulmonary fibrosis (RIPF) is a chronic, progressive, irreversible lung interstitial disease that develops after radiotherapy. Although several previous studies have focused on the mechanism of epithelial-mesenchymal transition (EMT) in lung epithelial cells, the essential factors involved in this process remain poorly understood. The DNA-dependent protein kinase catalytic subunit (DNA-PKcs) exhibits strong repair capacity when cells undergo radiation-induced damage; whether DNA-PKcs regulates EMT during RIPF remains unclear. OBJECTIVES: To investigate the role and molecular mechanism of DNA-PKcs in RIPF and provide an important theoretical basis for utilising DNA-PKcs-targeted drugs for preventing RIPF. METHODS: DNA-PKcs knockout (DPK-/-) mice were generated via the Cas9/sgRNA technique and subjected to whole chest ionizing radiation (IR) at a 20 Gy dose. Before whole chest IR, the mice were intragastrically administered the DNA-PKcs-targeted drug VND3207. Lung tissues were collected at 1 and 5 months after IR. RESULTS: The expression of DNA-PKcs is low in pulmonary fibrosis (PF) patients. DNA-PKcs deficiency significantly exacerbated RIPF by promoting EMT in lung epithelial cells. Mechanistically, DNA-PKcs deletion by shRNA or inhibitor NU7441 maintained the protein stability of Twist1. Furthermore, AKT1 mediated the interaction between DNA-PKcs and Twist1. High Twist1 expression and EMT-associated changes caused by DNA-PKcs deletion were blocked by insulin-like growth factor-1 (IGF-1), an AKT1 agonist. The radioprotective drug VND3207 prevented IR-induced EMT and alleviated RIPF in mice by stimulating the kinase activity of DNA-PKcs. CONCLUSION: Our study clarified the critical role and mechanism of DNA-PKcs in RIPF and showed that it could be a potential target for preventing RIPF.

Wnt signaling: Modulating tumor-associated macrophages and related immunotherapeutic insights.

Wnt signaling pathways are highly conserved cascades that mediate multiple biological processes through canonical or noncanonical pathways, from embryonic development to tissue maintenance, but they also contribute to the pathogenesis of numerous cancers. Recent studies have revealed that Wnt signaling pathways critically control the interplay between cancer cells and tumor-associated macrophages (TAMs) in the tumor microenvironment (TME) and potentially impact the efficacy of cancer immunotherapy. In this review, we summarize the evidence that Wnt signaling pathways boost the maturation and infiltration of macrophages for immune surveillance in the steady state but also polarize TAMs toward immunosuppressive M2-like phenotypes for immune escape in the TME. Both cancer cells and TAMs utilize Wnt signaling to transmit signals, and this interaction is crucial for the carcinogenesis and progression of common solid cancers, such as colorectal, gastric, hepatocellular, breast, thyroid, prostate, kidney, and lung cancers; osteosarcoma; and glioma. Specifically, compared with those in solid cancers, Wnt signaling pathways play a distinct role in the pathogenesis of leukemia. Efforts to develop Wnt-based drugs for cancer treatment are still ongoing, and some indeed enhance the anticancer immune response. We believe that the combination of Wnt signaling-based therapy with conventional or immune therapies is a promising therapeutic approach and can facilitate personalized treatment for most cancers.

Assessing bioartificial organ function: the 3P model framework and its validation.

The rapid advancement in the fabrication and culture of in vitro organs has marked a new era in biomedical research. While strides have been made in creating structurally diverse bioartificial organs, such as the liver, which serves as the focal organ in our study, the field lacks a uniform approach for the predictive assessment of liver function. Our research bridges this gap with the introduction of a novel, machine-learning-based "3P model" framework. This model draws on a decade of experimental data across diverse culture platform studies, aiming to identify critical fabrication parameters affecting liver function, particularly in terms of albumin and urea secretion. Through meticulous statistical analysis, we evaluated the functional sustainability of the in vitro liver models. Despite the diversity of research methodologies and the consequent scarcity of standardized data, our regression model effectively captures the patterns observed in experimental findings. The insights gleaned from our study shed light on optimizing culture conditions and advance the evaluation of the functional maintenance capacity of bioartificial livers. This sets a precedent for future functional evaluations of bioartificial organs using machine learning.

Investigating the impact of sebum secretion on liquid makeup foundation darkening and its solution.

OBJECTIVES: Darkening has been an issue of concern for foundation products. The secretion of sebum plays a significant role in the process of foundation darkening, but the underlying mechanisms and solutions have been rarely reported. The aim of this study was to explore the relationship between sebum secretion and liquid foundation darkening and to provide possible solutions for reducing sebum-induced darkening in liquid foundation. METHODS: Artificial sebum in different concentrations was added to a basic liquid foundation to simulate different stages of sebum secretion. The colour of the mixture was then measured by a spectrophotometer on the standard opacity chart. Potential technical solutions for anti-darkening were applied to a basic liquid foundation, and its ability to anti-darkening was further verified in vivo. RESULTS: (1) The influences of sebum addition on liquid foundation darkening had a significant positive correlation with the increase in transmissivities (R2 = 0.852, p < 0.01). (2) A certain range of sebum addition can reduce the darkening of volatile foundations. (3) The liquid foundations using pigments with high dispersibility in sebum were less influenced by sebum. (4) The replacement of pigments with oil-fixing ability could effectively reduce the darkening of liquid foundations induced by sebum (p < 0.01). CONCLUSION: The effect of sebum on the darkening of liquid foundation was accompanied by a greater transmissivity as its pigment concentration decreased. Balanced volatility, the addition of powders with higher sebum dispersibility and the replacement of oil-fixing powders could reduce the darkening of the liquid foundation caused by sebum secretion.

OBJECTIFS: L'assombrissement a été un problème de préoccupation pour les produits de fond de teint. La sécrétion de sébum joue un rôle significatif dans le processus d'assombrissement du fond de teint, mais les mécanismes sous-jacents et les solutions ont été rarement rapportés. L'objectif de cette étude était d'explorer la relation entre la sécrétion de sébum et l'assombrissement du fond de teint liquide, et de fournir des solutions possibles pour réduire l'assombrissement induit par le sébum dans le fond de teint liquide. MÉTHODES: Du sébum artificiel à différentes concentrations a été ajouté à un fond de teint liquide de base pour simuler différents stades de sécrétion de sébum. La couleur du mélange a ensuite été mesurée par un spectrophotomètre sur le tableau standard d'opacité. Des solutions techniques potentielles pour l'anti-assombrissement ont été appliquées à un fond de teint liquide de base et leur capacité à prévenir l'assombrissement a été vérifiée in vivo. RÉSULTATS: (1) Les influences de l'ajout de sébum sur l'assombrissement du fond de teint liquide avaient une corrélation significativement positive avec l'augmentation des transmissivités (R2 = 0.852 p < 0.01). (2) Une certaine plage de concentration de sebum peut réduire l'assombrissement des fondations volatiles. (3) Les fonds de teint liquides utilisant des pigments à haute dispersibilité dans le sébum étaient moins influencés par le sébum. (4) Le remplacement des pigments par des poudres à capacité de fixation d'huile pouvait efficacement réduire l'assombrissement des fonds de teint liquides induit par le sébum (p < 0.01). CONCLUSION: L'effet du sébum sur l'assombrissement du fond de teint liquide était accompagné d'une plus grande transmissivité à mesure que la concentration de son pigment diminuait. La volatilité équilibrée, l'ajout de poudres à plus grande dispersibilité de sébum et le remplacement de poudres à capacité de fixation d'huile pourraient réduire l'assombrissement du fond de teint liquide causé par la sécrétion de sébum.

MLatom 3: A Platform for Machine Learning-Enhanced Computational Chemistry Simulations and Workflows.

Machine learning (ML) is increasingly becoming a common tool in computational chemistry. At the same time, the rapid development of ML methods requires a flexible software framework for designing custom workflows. MLatom 3 is a program package designed to leverage the power of ML to enhance typical computational chemistry simulations and to create complex workflows. This open-source package provides plenty of choice to the users who can run simulations with the command-line options, input files, or with scripts using MLatom as a Python package, both on their computers and on the online XACS cloud computing service at XACScloud.com. Computational chemists can calculate energies and thermochemical properties, optimize geometries, run molecular and quantum dynamics, and simulate (ro)vibrational, one-photon UV/vis absorption, and two-photon absorption spectra with ML, quantum mechanical, and combined models. The users can choose from an extensive library of methods containing pretrained ML models and quantum mechanical approximations such as AIQM1 approaching coupled-cluster accuracy. The developers can build their own models using various ML algorithms. The great flexibility of MLatom is largely due to the extensive use of the interfaces to many state-of-the-art software packages and libraries.

Influence of landscape patterns on nitrate and particulate organic nitrogen inputs to urban stormwater runoff.

This study investigated the effect of the landscape pattern of permeable/impermeable patches on NO3--N and particulate organic nitrogen (PON) concentrations during stormwater runoff transport and their source contributions. Six landscape pattern indices, namely, mean proximity index (MPI), largest patch index (LPI), mean shape index (MSI), landscape shape index (LSI), connect index (CONNECT), and splitting index (SPLIT), were selected to reflect the fragmentation, complexity, and connectivity of permeable patches in urban catchments. The results show that lower fragmentation, higher complexity, and greater connectivity can reduce NO3--N concentrations in road runoff and drainage flow (i.e., the flow in the stormwater drainage network), as well as PON concentrations in road runoff. Further, the above landscape pattern is effective for mitigating the contributions of NO3--N and PON from road runoff. Low impact development (LID) can be incorporated with the landscape pattern of permeable/impermeable patches to mitigate nitrogen pollution in urban stormwater at the catchment scale by optimizing the spatial arrangement.

Energy-conserving molecular dynamics is not energy conserving.

Molecular dynamics (MD) is a widely-used tool for simulating molecular and materials properties. It is common wisdom that molecular dynamics simulations should obey physical laws and, hence, lots of effort is put into ensuring that molecular dynamics simulations are energy conserving. The emergence of machine learning (ML) potentials for MD leads to a growing realization that monitoring conservation of energy during simulations is of low utility because the dynamics is often unphysically dissociative. Other ML methods for MD are not based on a potential and provide only forces or trajectories which are reasonable but not necessarily energy-conserving. Here we propose to clearly distinguish between the simulation-energy and true-energy conservation and highlight that the simulations should focus on decreasing the degree of true-energy non-conservation. We introduce very simple, new criteria for evaluating the quality of molecular dynamics by estimating the degree of true-energy non-conservation and we demonstrate their practical utility on an example of infrared spectra simulations. These criteria are more important and intuitive than simply evaluating the quality of the ML potential energies and forces as is commonly done and can be applied universally, e.g., even for trajectories with unknown or discontinuous potential energy. Such an approach introduces new standards for evaluating MD by focusing on the true-energy conservation and can help in developing more accurate methods for simulating molecular and materials properties.

Four-Dimensional-Spacetime Atomistic Artificial Intelligence Models.

We demonstrate that AI can learn atomistic systems in the four-dimensional (4D) spacetime. For this, we introduce the 4D-spacetime GICnet model, which for the given initial conditions (nuclear positions and velocities at time zero) can predict nuclear positions and velocities as a continuous function of time up to the distant future. Such models of molecules can be unrolled in the time dimension to yield long-time high-resolution molecular dynamics trajectories with high efficiency and accuracy. 4D-spacetime models can make predictions for different times in any order and do not need a stepwise evaluation of forces and integration of the equations of motions at discretized time steps, which is a major advance over traditional, cost-inefficient molecular dynamics. These models can be used to speed up dynamics, simulate vibrational spectra, and obtain deeper insight into nuclear motions, as we demonstrate for a series of organic molecules.

CUDC-101 as a dual-target inhibitor of EGFR and HDAC enhances the anti-myeloma effects of bortezomib by regulating G2/M cell cycle arrest. / EGFR和HDAC双靶点抑制剂CUDC-101通过调控G2/M期阻滞增强硼替佐米抗骨髓瘤的作用.

CUDC-101, an effective and multi-target inhibitor of epidermal growth factor receptor (EGFR), histone deacetylase (HDAC), and human epidermal growth factor receptor 2 (HER2), has been reported to inhibit many kinds of cancers, such as acute promyelocytic leukemia and non-Hodgkin's lymphoma. However, no studies have yet investigated whether CUDC-101 is effective against myeloma. Herein, we proved that CUDC-101 effectively inhibits the proliferation of multiple myeloma (MM) cell lines and induces cell apoptosis in a time- and dose-dependent manner. Moreover, CUDC-101 markedly blocked the signaling pathway of EGFR/phosphoinositide-3-kinase (PI3K) and HDAC, and regulated the cell cycle G2/M arrest. Moreover, we revealed through in vivo experiment that CUDC-101 is a potent anti-myeloma drug. Bortezomib is one of the important drugs in MM treatment, and we investigated whether CUDC-101 has a synergistic or additive effect with bortezomib. The results showed that this drug combination had a synergistic anti-myeloma effect by inducing G2/M phase blockade. Collectively, our findings revealed that CUDC-101 could act on its own or in conjunction with bortezomib, which provides insights into exploring new strategies for MM treatment.

[sEMG Wireless Acquisition System Based on CC3200 and ADS1299].

A multi-channel surface electromyography wireless acquisition system is designed, which is mainly composed of ADS1299 integrated analog front-end chip and CC3200 wireless MCU of TI company. The key indicators of hardware are measured according to the industry standard, and the results are better than the industry standard, which can meet the continuous use of multi-scene tasks. This system has the advantages of high performance, low power consumption and small size. It has been applied to the detection of surface EMG signal in motion gesture recognition and has a good application value.

Explicit Learning of Derivatives with the KREG and pKREG Models on the Example of Accurate Representation of Molecular Potential Energy Surfaces.

The KREG and pKREG models were proven to enable accurate learning of multidimensional single-molecule surfaces of quantum chemical properties such as ground-state potential energies, excitation energies, and oscillator strengths. These models are based on kernel ridge regression (KRR) with the Gaussian kernel function and employ a relative-to-equilibrium (RE) global molecular descriptor, while pKREG is designed to enforce invariance under atom permutations with a permutationally invariant kernel. Here we extend these two models to also explicitly include the derivative information from the training data into the models, which greatly improves their accuracy. We demonstrate on the example of learning potential energies and energy gradients that KREG and pKREG models are better or on par with state-of-the-art machine learning models. We also found that in challenging cases both energy and energy gradient labels should be learned to properly model potential energy surfaces and learning only energies or gradients is insufficient. The models' open-source implementation is freely available in the MLatom package for general-purpose atomistic machine learning simulations, which can be also performed on the MLatom@XACS cloud computing service.