From data to noise to data for mixing physics across temperatures with generative artificial intelligence.

Using simulations or experiments performed at some set of temperatures to learn about the physics or chemistry at some other arbitrary temperature is a problem of immense practical and theoretical relevance. Here we develop a framework based on statistical mechanics and generative artificial intelligence that allows solving this problem. Specifically, we work with denoising diffusion probabilistic models and show how these models in combination with replica exchange molecular dynamics achieve superior sampling of the biomolecular energy landscape at temperatures that were never simulated without assuming any particular slow degrees of freedom. The key idea is to treat the temperature as a fluctuating random variable and not a control parameter as is usually done. This allows us to directly sample from the joint probability distribution in configuration and temperature space. The results here are demonstrated for a chirally symmetric peptide and single-strand RNA undergoing conformational transitions in all-atom water. We demonstrate how we can discover transition states and metastable states that were previously unseen at the temperature of interest and even bypass the need to perform further simulations for a wide range of temperatures. At the same time, any unphysical states are easily identifiable through very low Boltzmann weights. The procedure while shown here for a class of molecular simulations should be more generally applicable to mixing information across simulations and experiments with varying control parameters.

Research Review: Grandparental care and child mental health - a systematic review and meta-analysis.

BACKGROUND: The number of children residing in grandfamilies is growing worldwide, leading to more research attention on grandparental care over the past decades. Grandparental care can influence child well-being in various forms and the effects vary across contexts. In this systematic review and meta-analysis, we synthesize the evidence on the relation between grandparental care and children's mental health status. METHODS: We identified 5,745 records from seven databases, among which 38 articles were included for review. Random effects meta-analyses were used to synthesize evidence from eligible studies. We also examined the variability across study and participant characteristics, including study design, recruitment method, child age, child gender, study region, family type, comparison group, and outcome rater. RESULTS: The meta-analysis consisted of 344,860 children from the included studies, whose average age was 10.29, and of which 51.39% were female. Compared with their counterparts, children being cared for by their grandparents had worse mental health status, including more internalizing problems (d = -0.20, 95% CI [-0.31, -0.09], p = .001), externalizing problems (d = -0.11, 95% CI [-0.21, -0.01], p = .03), overall mental problems (d = -0.37, 95% CI [-0.70, -0.04], p = .03), and poorer socioemotional well-being (d = -0.26, 95% CI [-0.49, -0.03], p = .03). The effects varied by study design and child gender. CONCLUSIONS: The findings highlight that grandparental care is negatively associated with child mental health outcomes with trivial-to-small effect sizes. More supportive programs and interventions should be delivered to grandfamilies, especially in disadvantaged communities.

Old trees bloom new flowers, lysosome targeted near-infrared fluorescent probe for ratiometric sensing of hypobromous acid in vitro and in vivo based on Nile red skeleton.

Hypobromous acid (HOBr), one of the significant reactive oxygen species (ROS) that acts as an important role in human immune system, however the increasing level of HOBr in human body can cause the disorder of eosinophils (EPO), leading to oxidative stress in organelles, and further causing a series of diseases. In this study, a ratiometric fluorescent probe DMBP based on Nile red skeleton was developed to detect HOBr specifically by the electrophilic substitution with HOBr. DMBP emits near-infrared (NIR) fluorescence at 653 nm, after reacting with HOBr, the emission wavelength of DMBP shifted blue and a new peak appeared at 520 nm, realizing a ratiometric examination of HOBr with a limit of detection of 89.00 nM. Based on its sensitive and specific response to HOBr, DMBP was applied in the visual imaging of HOBr in HepG2 cells and zebrafish. Foremost, probe DMBP has excellent lysosome targeting ability and NIR emission reduced the background interference of biological tissues, providing a potential analytical tool to further investigate the role of HOBr in lysosome.

Biological impact and therapeutic potential of a novel camptothecin derivative (FLQY2) in pancreatic cancer through inactivation of the PDK1/AKT/mTOR pathway.

BACKGROUND: Camptothecin (CPT), a pentacyclic alkaloid with antitumor properties, is derived from the Camptotheca acuminata. Topotecan and irinotecan (CPT derivatives) were first approved by the Food and Drug Administration for cancer treatment over 25 years ago and remain key anticancer drugs today. However, their use is often limited by clinical toxicity. Despite extensive development efforts, many of these derivatives have not succeeded clinically, particularly in their effectiveness against pancreatic cancer which remains modest. AIM OF THE STUDY: This study aimed to evaluate the therapeutic activity of FLQY2, a CPT derivative synthesized in our laboratory, against pancreatic cancer, comparing its efficacy and mechanism of action with those of established clinical drugs. METHODS: The cytotoxic effects of FLQY2 on cancer cells were assessed using an MTT assay. Patient-derived organoid (PDO) models were employed to compare the sensitivity of FLQY2 to existing clinical drugs across various cancers. The impact of FLQY2 on apoptosis and cell cycle arrest in Mia Paca-2 pancreatic cancer cells was examined through flow cytometry. Transcriptomic and proteomic analyses were conducted to explore the underlying mechanisms of FLQY2's antitumor activity. Western blotting was used to determine the levels of proteins regulated by FLQY2. Additionally, the antitumor efficacy of FLQY2 in vivo was evaluated in a pancreatic cancer xenograft model. RESULTS: FLQY2 demonstrated (1) potent cytotoxicity; (2) superior tumor-suppressive activity in PDO models compared to current clinical drugs such as gemcitabine, 5-fluorouracil, cisplatin, paclitaxel, ivosidenib, infinitinib, and lenvatinib; (3) significantly greater tumor inhibition than paclitaxel liposomes in a pancreatic cancer xenograft model; (4) robust antitumor effects, closely associated with the inhibition of the TOP I and PDK1/AKT/mTOR signaling pathways. In vitro studies revealed that FLQY2 inhibited cell proliferation, colony formation, induced apoptosis, and caused cell cycle arrest at nanomolar concentrations. Furthermore, the combination of FLQY2 and gemcitabine exhibited significant inhibitory and synergistic effects. CONCLUSION: The study confirmed the involvement of topoisomerase I and the PDK1/AKT/mTOR pathways in mediating the antitumor activity of FLQY2 in treating Mia Paca-2 pancreatic cancer. Therefore, FLQY2 has potential as a novel therapeutic option for patients with pancreatic cancer.

TIRSF: a web server for screening gene signatures to predict Tumor immunotherapy response.

Immune checkpoint blockade (ICB) therapy has been successfully applied to clinically therapeutics in multiple cancers, but its efficacy varies greatly among different patients and cancer types. Therefore, the construction of gene signatures to identify patients who could benefit from ICB therapy is particularly important for precision cancer treatment. However, due to the lack of a user-friendly platform, the construction of such gene signatures is a great challenge for clinical investigators who have limited programming skills. In light of this challenge, we developed a web server called Tumor Immunotherapy Response Signature Finder(TIRSF) for the construction of gene signatures to predict ICB therapy response in cancer patients. TIRSF consists of three functional modules. The first module is the Signature Discovery module which provides signature construction and performance evaluation functionalities. The second is a module for response prediction based on the TIRSF signatures, which enables response prediction and prognostic analysis of immunotherapy samples. The last is a module for response prediction based on existing signatures. This module currently integrates 24 published signatures for ICB therapy response prediction. Together, all of above features can be freely accessed at http://tirsf.renlab.org/.

Parental Depression and Self-Stigma Among Chinese Young People Living With Depression: A Qualitative Study.

Self-stigma is detrimental to psychosocial well-being and the recovery journey among people living with depression. However, there has been limited research exploring the experience of stigma internalization when depression runs in families. This study aims to address this gap by (1) characterizing the manifestations of self-stigma among individuals living with depression whose parent(s) also have depression and (2) exploring the potential mechanisms underlying the impact of parental depression on self-stigma. Essential principles of the constructivist grounded theory approach were adopted to collect data through in-depth interviews with 27 participants aged 15-30, living in Mainland China. Many participants perceived depression running in their family as an endless disaster and an incurable illness. These beliefs further led to stigmatizing emotions (such as suppression, anger, and guilt) and behaviors (such as concealment and social withdrawal). Participants also highlighted ambivalent intergenerational relationships, tense family atmospheres, lower parental emotional involvement and support, and a lack of family flexibility due to parental depression. Furthermore, parental depression impacted participants' self-stigma by interfering with family relationships, family functioning, and parenting styles. It also shaped their perceptions of family, illness attribution, and public stigma. Additionally, parental depression had an impact on participants' social functioning, self-esteem, and personality, making them more susceptible to self-stigma. This study emphasizes the crucial role that the family plays in the internalization of stigma among individuals living with depression. It suggests that family dynamics, rather than family structure or economic backgrounds alone, shape this process.

[Evaluation of chemical constituents of Draconis Sanguis and its protective effect on renal tubular injury in diabetic kidney disease].

The chemical constituents of Draconis Sanguis were preliminarily studied by macroporous resin, silica gel, dextran gel, and high-performance liquid chromatography. One retro-dihydrochalcone, four flavonoids, and one stilbene were isolated. Their chemical structures were identified as 4-hydroxy-2,6-dimethoxy-3-methyldihydrochalcone(1), 4'-hydroxy-5,7-dimethoxy-8-methylflavan(2), 7-hydroxy-4',5-dimethoxyflavan(3),(2S)-7-hydroxy-5-methoxy-6-methylflavan(4),(2S)-7-hydroxy-5-methoxyflavan(5), and pterostilbene(6) by modern spectroscopy, physicochemical properties, and literature comparison. Compound 1 was a new compound. Compounds 2 and 6 were first found in the Arecaceae family. Compound 5 had the potential to prevent and treat diabetic kidney disease.

A DDoS Detection Method Based on Feature Engineering and Machine Learning in Software-Defined Networks.

Distributed denial-of-service (DDoS) attacks pose a significant cybersecurity threat to software-defined networks (SDNs). This paper proposes a feature-engineering- and machine-learning-based approach to detect DDoS attacks in SDNs. First, the CSE-CIC-IDS2018 dataset was cleaned and normalized, and the optimal feature subset was found using an improved binary grey wolf optimization algorithm. Next, the optimal feature subset was trained and tested in Random Forest (RF), Support Vector Machine (SVM), K-Nearest Neighbor (k-NN), Decision Tree, and XGBoost machine learning algorithms, from which the best classifier was selected for DDoS attack detection and deployed in the SDN controller. The results show that RF performs best when compared across several performance metrics (e.g., accuracy, precision, recall, F1 and AUC values). We also explore the comparison between different models and algorithms. The results show that our proposed method performed the best and can effectively detect and identify DDoS attacks in SDNs, providing a new idea and solution for the security of SDNs.

RUC-Net: A Residual-Unet-Based Convolutional Neural Network for Pixel-Level Pavement Crack Segmentation.

Automatic crack detection is always a challenging task due to the inherent complex backgrounds, uneven illumination, irregular patterns, and various types of noise interference. In this paper, we proposed a U-shaped encoder-decoder semantic segmentation network combining Unet and Resnet for pixel-level pavement crack image segmentation, which is called RUC-Net. We introduced the spatial-channel squeeze and excitation (scSE) attention module to improve the detection effect and used the focal loss function to deal with the class imbalance problem in the pavement crack segmentation task. We evaluated our methods using three public datasets, CFD, Crack500, and DeepCrack, and all achieved superior results to those of FCN, Unet, and SegNet. In addition, taking the CFD dataset as an example, we performed ablation studies and compared the differences of various scSE modules and their combinations in improving the performance of crack detection.

Melatonin inhibiting the survival of human gastric cancer cells under ER stress involving autophagy and Ras-Raf-MAPK signalling.

Melatonin exhibits antitumour activities in the treatment of many human cancers. In the present study, we aimed to improve the therapeutic potential of melatonin in gastric cancer. Our results confirmed that melatonin dose-dependently suppressed the proliferation and necrosis, and increased G0/G1 phase arrest, apoptosis, autophagy and endoplasmic reticulum (ER) stress. The Ras-Raf-MAPK signalling pathway was activated in cells after melatonin treatment. RNA-seq was performed and GSEA analysis further confirmed that many down-regulated genes in melatonin-treated cells were associated with proliferation. However, GSEA analysis also indicated that many pathways related to metastasis were increased after melatonin treatment. Subsequently, combinatorial treatment was conducted to further investigate the therapeutic outcomes of melatonin. A combination of melatonin and thapsigargin increased the apoptotic rate and G0/G1 cell cycle arrest when compared to treatment with melatonin alone. Melatonin in combination with thapsigargin triggered the increased expression of Bip, LC3-II, phospho-Erk1/2 and phospho-p38 MAPK. In addition, STF-083010, an IRE1a inhibitor, further exacerbated the decrease in survival rate induced by combinatorial treatment with melatonin and thapsigargin. Collectively, melatonin was effective in gastric cancer treatment by modifying ER stress.

ABC and ABAB Block Copolymers by Electrochemically Controlled Ring-Opening Polymerization.

An electrochemically controlled synthesis of multiblock copolymers by alternating the redox states of (salfan)Zr(OtBu)2 (salfan = 1,1'-di(2-tert-butyl-6-N-methylmethylenephenoxy)ferrocene) is reported. Aided by electrochemistry with a glassy carbon working electrode, an in situ potential switch alters the catalyst's oxidation state and its subsequent monomer (l-lactide, ß-butyrolactone, or cyclohexene oxide) selectivity in one pot. Various multiblock copolymers were prepared, including an ABAB tetrablock copolymer, poly(cyclohexene oxide-b-lactide-b-cyclohexene oxide-b-lactide), and an ABC triblock copolymer, poly(hydroxybutyrate-b-cyclohexene oxide-b-lactide). The polymers produced using this technique are similar to those produced via a chemical redox reagent method, displaying moderately narrow dispersities (1.1-1.5) and molecular weights ranging from 7 to 26 kDa.

Genome-wide identification of auxin response factor (ARF) family in kiwifruit (Actinidia chinensis) and analysis of their inducible involvements in abiotic stresses.

Auxin response factor (ARF) acts as a vital component of auxin signaling and participates in growth, development, and stress responses in plants. In the present study, we comprehensively analyzed kiwifruit's (Actinidia chinensis) ARF genes (AcARFs) and their involvement in abiotic stress response. We identified a total of 41 AcARFs encoding ARFs in the A. chinensis genome. AcARF genes were characterized by the classic ARF_resp and a B3 domain and primarily localized on the cytoplasm and nucleus. AcARFs were categorized into eight subgroups as per the phylogenetic analysis. Synteny analysis showed that 35 gene pairs in AcARF family underwent segmental and whole genome duplication events. Promoter cis-element prediction revealed that AcARFs might be involved in abiotic factors related to stress response, which was later assessed and validated by qRT-PCR based expression analysis. Additionally, AcARFs showed tissue-specific expression. These findings extend our understanding of the functional roles of AcARFs in stress responses. Taken together, the systematic annotation of the AcARF family genes provides a platform for the functional and evolutionary study, which might help in elucidating the precise roles of the AcARFs in stress responses. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1007/s12298-021-01011-4.

Understanding the role of predictive time delay and biased propagator in RAVE.

In this work, we revisit our recent iterative machine learning (ML)-molecular dynamics (MD) technique "Reweighted autoencoded variational Bayes for enhanced sampling" [J. M. L. Ribeiro et al., J. Chem. Phys. 149, 072301 (2018) and Y. Wang, J. M. L. Ribeiro, and P. Tiwary, Nat. Commun. 10, 3573 (2019)] and analyze and formalize some of its approximations. These include (a) the choice of a predictive time-delay, or how far into the future should the ML try to predict the state of a given system output from MD, and (b) that for short time-delays, how much of an error is made in approximating the biased propagator for the dynamics as the unbiased propagator. We demonstrate through a master equation framework as to why the exact choice of time-delay is irrelevant as long as a small non-zero value is adopted. We also derive a correction to reweight the biased propagator, and somewhat to our dissatisfaction but also to our reassurance, we find that it barely makes a difference to the intuitive picture we had previously derived and used.

Confronting pitfalls of AI-augmented molecular dynamics using statistical physics.

Artificial intelligence (AI)-based approaches have had indubitable impact across the sciences through the ability to extract relevant information from raw data. Recently, AI has also found use in enhancing the efficiency of molecular simulations, wherein AI derived slow modes are used to accelerate the simulation in targeted ways. However, while typical fields where AI is used are characterized by a plethora of data, molecular simulations, per construction, suffer from limited sampling and thus limited data. As such, the use of AI in molecular simulations can suffer from a dangerous situation where the AI-optimization could get stuck in spurious regimes, leading to incorrect characterization of the reaction coordinate (RC) for the problem at hand. When such an incorrect RC is then used to perform additional simulations, one could start to deviate progressively from the ground truth. To deal with this problem of spurious AI-solutions, here, we report a novel and automated algorithm using ideas from statistical mechanics. It is based on the notion that a more reliable AI-solution will be one that maximizes the timescale separation between slow and fast processes. To learn this timescale separation even from limited data, we use a maximum caliber-based framework. We show the applicability of this automatic protocol for three classic benchmark problems, namely, the conformational dynamics of a model peptide, ligand-unbinding from a protein, and folding/unfolding energy landscape of the C-terminal domain of protein G. We believe that our work will lead to increased and robust use of trustworthy AI in molecular simulations of complex systems.

A Digital Closed-Loop Sense MEMS Disk Resonator Gyroscope Circuit Design Based on Integrated Analog Front-end.

A digital closed-loop system design of a microelectromechanical systems (MEMS) disk resonator gyroscope (DRG) is proposed in this paper. Vibration models with non-ideal factors are provided based on the structure characteristics and operation mode of the sensing element. The DRG operates in force balance mode with four control loops. A closed self-excited loop realizes stable vibration amplitude on the basis of peak detection technology and phase control loop. Force-to-rebalance technology is employed for the closed sense loop. A high-frequency carrier loaded on an anchor weakens the effect of parasitic capacitances coupling. The signal detected by the charge amplifier is demodulated and converted into a digital output for subsequent processing. Considering compatibility with digital circuits and output precision demands, a low passband sigma-delta (ΣΔ) analog-to-digital converter (ADC) is implemented with a 111.8dB signal-to-noise ratio (SNR). The analog front-end and digital closed self-excited loop is manufactured with a standard 0.35 µm complementary metal-oxide-semiconductor (CMOS) technology. The experimental results show a bias instability of 2.1 °/h and a nonlinearity of 0.035% over the ± 400° full-scale range.

Kinetics of Ligand-Protein Dissociation from All-Atom Simulations: Are We There Yet?

Large parallel gains in the development of both computational resources and sampling methods have now made it possible to simulate dissociation events in ligand-protein complexes with all-atom resolution. Such encouraging progress, together with the inherent spatiotemporal resolution associated with molecular simulations, has left their use for investigating dissociation processes brimming with potential, both in rational drug design, where it can be an invaluable tool for determining the mechanistic driving forces behind dissociation rate constants, and in force-field development, where it can provide a catalog of transient molecular structures with which to refine force fields. Although much progress has been made in making force fields more accurate, reducing their error for transient structures along a transition path could yet prove to be a critical development helping to make kinetic predictions much more accurate. In what follows, we will provide a state-of-the-art compilation of the enhanced sampling methods based on molecular dynamics (MD) simulations used to investigate the kinetics and mechanisms of ligand-protein dissociation processes. Due to the time scales of such processes being slower than what is accessible using straightforward MD simulations, several ingenious schemes are being devised at a rapid rate to overcome this obstacle. Here we provide an up-to-date compendium of such methods and their achievements and shortcomings in extracting mechanistic insight into ligand-protein dissociation. We conclude with a critical and provocative appraisal attempting to answer the title of this Perspective.

The transcriptomic responses of the ark shell, Anadara broughtonii, to sulfide and hypoxia exposure.

Sulfide and hypoxia threaten marine organisms in various ways. Anadara broughtonii, a commercial marine bivalve in China which has great potential exposure to sulfide and hypoxia, was selected to test the responses to these stresses. Digital gene expression profile (DGE) analysis was performed on the juveniles' gills after exposed to normal condition (CG group), hypoxia (LO group), and low/high concentration of sulfide (LS/HS group, administered in hypoxia), respectively, using RNA-seq technology. A total of over 30 million clean reads were filtered from each DGE library and over 90% of them were annotated successfully. In total, 774 significant differentially expressed genes (DEGs) were detected and assigned to Gene ontology (GO) classification and KEGG Pathway enrichment analysis. The results show that many of the upregulated DEGs are related to hemoglobin, immunology, and stress responding. In the stressed A. broughtonii, cytochrome P450 and phosphoenolpyruvate carboxykinase may stimulate the glycolysis process to reduce oxygen consumption; Aminoacyl-tRNA synthetases, heat shock protein and protein disulfide isomerase probably help to maintain the genome integrity; Baculoviral IAP repeat-containing protein 2/3, mitogen-activated protein kinase and tumor necrosis factor pathways were probably responsible for protein repair, proteolysis, apoptosis and immune responses to high concentration of sulfide. Combined challenges also induced alternative oxidase and sushi repeat-containing protein, which have indistinct but probably indispensable function in invertebrates. For the first time, comprehensive transcriptome information on A. broughtonii in response to sulfide and hypoxia were provided. Our research offers new insights into the molecular mechanism behind the resistance of shellfish to sulfide and hypoxia.

A Straightforward Approach for Synthesizing Electromechanical Sigma-Delta MEMS Accelerometers.

The EM- Σ Δ (electromechanical sigma-delta) approach is a concise and efficient way to realize the digital interface for micro-electromechanical systems (MEMS) accelerometers. However, including a fixed MEMS element makes the synthesizing of the EM- Σ Δ loop an intricate problem. The loop parameters of EM- Σ Δ can not be directly mapped from existing electrical Σ Δ modulator, and the synthesizing problem relies an experience-dependent trail-and-error procedure. In this paper, we provide a new point of view to consider the EM- Σ Δ loop. The EM- Σ Δ loop is analyzed in detail from aspects of the signal loop, displacement modulation path and digital quantization loop. By taking a separate consideration of the signal loop and quantization noise loop, the design strategy is made clear and straightforward. On this basis, a discrete-time PID (proportional integral differential) loop compensator is introduced which enhances the in-band loop gain and suppresses the displacement modulation path, and hence, achieves better performance in system linearity and stability. A fifth-order EM- Σ Δ accelerometer system was designed and fabricated using 0.35 µ m CMOS-BCD technology. Based on proposed architecture and synthesizing procedure, the design effort was saved, and the in-band performance, linearity and stability were improved. A noise floor of 1 µ g / Hz , with a bandwidth 1 kHz and a dynamic range of 140 dB was achieved.

Acyl-CoA-binding protein family members in laticifers are possibly involved in lipid and latex metabolism of Hevea brasiliensis (the Para rubber tree).

BACKGROUND: Acyl-CoA-binding proteins (ACBPs) are mainly involved in acyl-CoA ester binding and trafficking in eukaryotic cells, and their various functions have been characterized in model plants, such as Arabidopsis thaliana (A. thaliana), Oryza sativa (rice), and other plant species. In the present study, genome-wide mining and expression analysis of ACBP genes was performed on Hevea brasiliensis (the para rubber tree), the most important latex-producing crop in the world. RESULTS: Six members of the H. brasiliensis ACBP family genes, designated HbACBP1-HbACBP6, were identified from the H. brasiliensis genome. They can be categorized into four classes with different amino acid sequences and domain structures based on the categorization of their A. thaliana counterparts. Phylogenetic analysis shows that the HbACBPs were clustered with those of other closely related species, such as Manihot esculenta, Ricinus communis, and Jatropha carcas, but were further from those of A. thaliana, a distantly related species. Expression analysis demonstrated that the HbACBP1 and HbACBP2 genes are more prominently expressed in H. brasiliensis latex, and their expression can be significantly enhanced by bark tapping (a mechanical wound) and jasmonic acid stimulation, whereas HbACBP3-HbACBP6 had almost the same expression patterns with relatively high levels in mature leaves and male flowers, but a markedly low abundance in the latex. HbACBP1 and HbACBP2 may have crucial roles in lipid and latex metabolism in laticifers, so their subcellular location was further investigated and the results indicated that HbACBP1 is a cytosol protein, whereas HbACBP2 is an endoplasmic reticulum-associated ACBP. CONCLUSIONS: In this study, the H. brasiliensis ACBP family genes were identified. Phylogenetic analyses of the HbABCPs indicate that there is a high conservation and evolutionary relationship between ACBPs in land plants. The HbACBPs are organ/tissue-specifically expressed and have different expression patterns in response to stimulation by bark tapping or ethrel/jasmonic acid. HbACBP1 and HbACBP2 are two important latex ACBPs that might be involved in the lipid and latex metabolism. The results may provide valuable information for further investigations into the biological functions of HbACBPs during latex metabolism and stress responses in H. brasiliensis.

Reweighted autoencoded variational Bayes for enhanced sampling (RAVE).

Here we propose the reweighted autoencoded variational Bayes for enhanced sampling (RAVE) method, a new iterative scheme that uses the deep learning framework of variational autoencoders to enhance sampling in molecular simulations. RAVE involves iterations between molecular simulations and deep learning in order to produce an increasingly accurate probability distribution along a low-dimensional latent space that captures the key features of the molecular simulation trajectory. Using the Kullback-Leibler divergence between this latent space distribution and the distribution of various trial reaction coordinates sampled from the molecular simulation, RAVE determines an optimum, yet nonetheless physically interpretable, reaction coordinate and optimum probability distribution. Both then directly serve as the biasing protocol for a new biased simulation, which is once again fed into the deep learning module with appropriate weights accounting for the bias, the procedure continuing until estimates of desirable thermodynamic observables are converged. Unlike recent methods using deep learning for enhanced sampling purposes, RAVE stands out in that (a) it naturally produces a physically interpretable reaction coordinate, (b) is independent of existing enhanced sampling protocols to enhance the fluctuations along the latent space identified via deep learning, and (c) it provides the ability to easily filter out spurious solutions learned by the deep learning procedure. The usefulness and reliability of RAVE is demonstrated by applying it to model potentials of increasing complexity, including computation of the binding free energy profile for a hydrophobic ligand-substrate system in explicit water with dissociation time of more than 3 min, in computer time at least twenty times less than that needed for umbrella sampling or metadynamics.