Deep Generative Models Research Articles

Discovery of Potential Neonicotinoid Insecticides by an Artificial Intelligence Generative Model and Structure-Based Virtual Screening.

The identification of neonicotinoid insecticides bearing novel scaffolds is of great importance for pesticide discovery. Here, artificial intelligence-based tools and virtual screening strategy were integrated to discover potential leads of neonicotinoid insecticides. A deep generative model was successfully constructed using a recurrent neural network combined with transfer learning. The model evaluation showed that the pretrained model could accurately grasp the SMILES grammar of drug-like molecules and generate potential neonicotinoid compounds after transfer learning. The generated molecules were evaluated by hierarchical virtual screening, hits were subjected to a similarity search, and the most similar structures were purchased for the bioassay. Compounds A2 and A5 displayed 52.5 and 50.3% mortality rates against Aphis craccivora at 100 mg/L, respectively. The docking study indicated that these two compounds have similar binding modes to neonicotinoids, which were verified by further molecular dynamics simulations.

Microstructure reconstruction of 2D/3D random materials via diffusion-based deep generative models

Microstructure reconstruction serves as a crucial foundation for establishing process–structure–property (PSP) relationship in material design. Confronting the limitations of variational autoencoder and generative adversarial network within generative models, this study adopted the denoising diffusion probabilistic model (DDPM) to learn the probability distribution of high-dimensional raw data and successfully reconstructed the microstructures of various composite materials, such as inclusion materials, spinodal decomposition materials, chessboard materials, fractal noise materials, and so on. The quality of generated microstructure was evaluated using quantitative measures like spatial correlation functions and Fourier descriptor. On this basis, this study also achieved the regulation of microstructure randomness and the generation of gradient materials through continuous interpolation in latent space using denoising diffusion implicit model (DDIM). Furthermore, the two-dimensional microstructure reconstruction was extended to three-dimensional framework and integrated permeability as a feature encoding embedding. This enables the conditional generation of three-dimensional microstructures for random porous materials within a defined permeability range. The permeabilities of these generated microstructures were further validated through the application of the lattice Boltzmann method. The above methods provide new ideas and references for material reverse design.

Disentangled deep generative models reveal coding principles of the human face processing network.

Despite decades of research, much is still unknown about the computations carried out in the human face processing network. Recently, deep networks have been proposed as a computational account of human visual processing, but while they provide a good match to neural data throughout visual cortex, they lack interpretability. We introduce a method for interpreting brain activity using a new class of deep generative models, disentangled representation learning models, which learn a low-dimensional latent space that "disentangles" different semantically meaningful dimensions of faces, such as rotation, lighting, or hairstyle, in an unsupervised manner by enforcing statistical independence between dimensions. We find that the majority of our model's learned latent dimensions are interpretable by human raters. Further, these latent dimensions serve as a good encoding model for human fMRI data. We next investigate the representation of different latent dimensions across face-selective voxels. We find that low- and high-level face features are represented in posterior and anterior face-selective regions, respectively, corroborating prior models of human face recognition. Interestingly, though, we find identity-relevant and irrelevant face features across the face processing network. Finally, we provide new insight into the few "entangled" (uninterpretable) dimensions in our model by showing that they match responses in the ventral stream and carry information about facial identity. Disentangled face encoding models provide an exciting alternative to standard "black box" deep learning approaches for modeling and interpreting human brain data.

Development of a deep-learning model for phase-separation structure of diblock copolymer based on self-consistent field analysis

Self-consistent field (SCF) analysis is an indispensable tool for predicting the microphase separation structures of polymer alloys. However, the computation of the phase-separated structures in the equilibrium state is computationally intensive, leading to high costs. To address this challenge, we propose a novel deep learning approach that leverages a generative adversarial network (GAN), a powerful deep generative model, to accelerate SCF analysis. Specifically, we trained the GAN using comprehensive data obtained from SCF analysis, enabling us to generate various images of feasible structures that are similar to the SCF analysis results. Our results demonstrate that the latent variables in the GAN are linked to the physical parameters and features of the phase-separation structures.

Research on Parkinson's disease recognition algorithm based on sample enhancement

Parkinson's disease patients have early vocal cord damage, and their voiceprint characteristics differ significantly from those of healthy individuals, which can be used to identify Parkinson's disease. However, the samples of the voiceprint dataset of Parkinson's disease patients are insufficient, so this paper proposes a double self-attention deep convolutional generative adversarial network model for sample enhancement to generate high-resolution spectrograms, based on which deep learning is used to recognize Parkinson's disease. This model improves the texture clarity of samples by increasing network depth and combining gradient penalty and spectral normalization techniques, and a family of pure convolutional neural networks (ConvNeXt) classification network based on Transfer learning is constructed to extract voiceprint features and classify them, which improves the accuracy of Parkinson's disease recognition. The validation experiments of the effectiveness of this paper's algorithm are carried out on the Parkinson's disease speech dataset. Compared with the pre-sample enhancement, the clarity of the samples generated by the proposed model in this paper as well as the Fréchet inception distance (FID) are improved, and the network model in this paper is able to achieve an accuracy of 98.8%. The results of this paper show that the Parkinson's disease recognition algorithm based on double self-attention deep convolutional generative adversarial network sample enhancement can accurately distinguish between healthy individuals and Parkinson's disease patients, which helps to solve the problem of insufficient samples for early recognition of voiceprint data in Parkinson's disease. In summary, the method effectively improves the classification accuracy of small-sample Parkinson's disease speech dataset and provides an effective solution idea for early Parkinson's disease speech diagnosis.

Nonlinear system identification using modified variational autoencoders

This research proposes a methodology for identifying nonlinear systems using input/output data and deep learning generative models. Our framework integrates Variational Autoencoders (VAE) with Nonlinear Autoregressive with exogenous input (NARX) in a unified identification structure to address overfitting in nonlinear system identification using NARX structures. Specifically, we modify a variational autoencoder by replacing the decoder module with a NARX model using the latent space information captured from the VAE encoder module as one of the exogenous inputs. Following the training phase, the decoder module can be used as a nonlinear model of the system. We evaluate the efficacy of our approach by performing open-loop prediction tests on data from four nonlinear benchmark systems: Cascaded tanks, Gas furnace, Silverbox, and Wiener-Hammerstein. The proposed VAE-NARX method reported Root Mean Squared Error (RMSE) of 8.23×10−3, 16.69×10−3, 0.002×10−3 and 0.037×10−3 respectively. Our results demonstrate that our proposed method achieves similar and outperforms prediction performances to standard identification techniques and can enhance the performance of traditional nonlinear system identification methods based on multi-layer perceptron models.

A review of deep learning‐based approaches for deepfake content detection

AbstractRecent advancements in deep learning generative models have raised concerns as they can create highly convincing counterfeit images and videos. This poses a threat to people's integrity and can lead to social instability. To address this issue, there is a pressing need to develop new computational models that can efficiently detect forged content and alert users to potential image and video manipulations. This paper presents a comprehensive review of recent studies for deepfake content detection using deep learning‐based approaches. We aim to broaden the state‐of‐the‐art research by systematically reviewing the different categories of fake content detection. Furthermore, we report the advantages and drawbacks of the examined works, and prescribe several future directions towards the issues and shortcomings still unsolved on deepfake detection.

De novo prediction of RNA 3D structures with deep generative models.

We present a Deep Learning approach to predict 3D folding structures of RNAs from their nucleic acid sequence. Our approach combines an autoregressive Deep Generative Model, Monte Carlo Tree Search, and a score model to find and rank the most likely folding structures for a given RNA sequence. We show that RNA de novo structure prediction by deep learning is possible at atom resolution, despite the low number of experimentally measured structures that can be used for training. We confirm the predictive power of our approach by achieving competitive results in a retrospective evaluation of the RNA-Puzzles prediction challenges, without using structural contact information from multiple sequence alignments or additional data from chemical probing experiments. Blind predictions for recent RNA-Puzzle challenges under the name "Dfold" further support the competitive performance of our approach.

State-specific protein–ligand complex structure prediction with a multiscale deep generative model

State-specific protein–ligand complex structure prediction with a multiscale deep generative model

Sampling the lattice Nambu-Goto string using Continuous Normalizing Flows

Effective String Theory (EST) represents a powerful non-perturbative approach to describe confinement in Yang-Mills theory that models the confining flux tube as a thin vibrating string. EST calculations are usually performed using the zeta-function regularization: however there are situations (for instance the study of the shape of the flux tube or of the higher order corrections beyond the Nambu-Goto EST) which involve observables that are too complex to be addressed in this way. In this paper we propose a numerical approach based on recent advances in machine learning methods to circumvent this problem. Using as a laboratory the Nambu-Goto string, we show that by using a new class of deep generative models called Continuous Normalizing Flows it is possible to obtain reliable numerical estimates of EST predictions.

Deep learning-based diffusion tensor image generation model: a proof-of-concept study

This study created an image-to-image translation model that synthesizes diffusion tensor images (DTI) from conventional diffusion weighted images, and validated the similarities between the original and synthetic DTI. Thirty-two healthy volunteers were prospectively recruited. DTI and DWI were obtained with six and three directions of the motion probing gradient (MPG), respectively. The identical imaging plane was paired for the image-to-image translation model that synthesized one direction of the MPG from DWI. This process was repeated six times in the respective MPG directions. Regions of interest (ROIs) in the lentiform nucleus, thalamus, posterior limb of the internal capsule, posterior thalamic radiation, and splenium of the corpus callosum were created and applied to maps derived from the original and synthetic DTI. The mean values and signal-to-noise ratio (SNR) of the original and synthetic maps for each ROI were compared. The Bland–Altman plot between the original and synthetic data was evaluated. Although the test dataset showed a larger standard deviation of all values and lower SNR in the synthetic data than in the original data, the Bland–Altman plots showed each plot localizing in a similar distribution. Synthetic DTI could be generated from conventional DWI with an image-to-image translation model.

DynamicBind: predicting ligand-specific protein-ligand complex structure with a deep equivariant generative model

While significant advances have been made in predicting static protein structures, the inherent dynamics of proteins, modulated by ligands, are crucial for understanding protein function and facilitating drug discovery. Traditional docking methods, frequently used in studying protein-ligand interactions, typically treat proteins as rigid. While molecular dynamics simulations can propose appropriate protein conformations, they’re computationally demanding due to rare transitions between biologically relevant equilibrium states. In this study, we present DynamicBind, a deep learning method that employs equivariant geometric diffusion networks to construct a smooth energy landscape, promoting efficient transitions between different equilibrium states. DynamicBind accurately recovers ligand-specific conformations from unbound protein structures without the need for holo-structures or extensive sampling. Remarkably, it demonstrates state-of-the-art performance in docking and virtual screening benchmarks. Our experiments reveal that DynamicBind can accommodate a wide range of large protein conformational changes and identify cryptic pockets in unseen protein targets. As a result, DynamicBind shows potential in accelerating the development of small molecules for previously undruggable targets and expanding the horizons of computational drug discovery.

Deep generative models in the quest for anticancer drugs: ways forward

Drug discovery is a costly and time-consuming process, especially because of the significant expenses connected with the high percentage of clinical trial failures. As such, there is a need for new paradigms enabling the optimization of the various stages, from hit identification to market approval. The upsurge in the use of artificial intelligence (AI) technologies and the advent of deep learning (DL) demonstrated a lot of promise in rethinking and redesigning the traditional pipelines in drug discovery, including de novo molecular design. In this regard, generative models have greatly impacted the de novo design of molecules with desired properties and are being increasingly integrated into real world drug discovery campaigns. Herein, we will briefly appraise recent case studies utilizing generative models for chemical structure generation in the area of anticancer drug discovery. Finally, we will analyze current challenges and limitations as well as the possible strategies to overcome them, outlining potential future directions to advance this exciting field.

Predicting protein conformational ensembles using deep generative models

Predicting protein conformational ensembles using deep generative models

A Missing Well-Logs Imputation Method Based on Conditional Denoising Diffusion Probabilistic Models

Summary Well logs comprise sequential data detailing the geological properties of formations at varying depths encountered during drilling. They are fundamental for various applications in the petroleum industry. However, acquired well logs often contain noise and missing data, which impedes their utility. To address this, numerous methods have been developed to impute missing components in well logs, ranging from traditional deterministic methods to modern data-driven models. Despite their effectiveness, these methods face several challenges. First, many are deterministic, lacking the ability to capture and represent the inherent uncertainties in the data. In addition, they often require complete logging data as input, which presents challenges in data sets with substantial missing data. Moreover, most are predictive models designed with specific targets that require retraining for different variables, which limits their versatility in handling data sets with diverse missing components. This work proposes the use of a generative model based on the conditional denoising diffusion probabilistic model (CDDPM) to impute missing components within well logs. The CDDPM offers several advantages. Its inherent probabilistic nature allows it to capture uncertainties in the data, providing predictions in the form of probability distributions rather than single-point estimates. This helps engineers make more robust and informed decisions in practice, thus mitigating potential risks. More importantly, due to its generative nature, the model is trained to learn the underlying data distribution, not the specific input-output map, which enables it to impute all missing data simultaneously. Through experiments on a real-world data set, we demonstrate that our proposed method surpasses conventional data-driven techniques in performance. Both qualitative and quantitative evaluations confirm the effectiveness of the model in imputing missing components. This research highlights the potential of modern deep generative models in petroleum engineering applications.

Emergent communication of multimodal deep generative models based on Metropolis-Hastings naming game.

Deep generative models (DGM) are increasingly employed in emergent communication systems. However, their application in multimodal data contexts is limited. This study proposes a novel model that combines multimodal DGM with the Metropolis-Hastings (MH) naming game, enabling two agents to focus jointly on a shared subject and develop common vocabularies. The model proves that it can handle multimodal data, even in cases of missing modalities. Integrating the MH naming game with multimodal variational autoencoders (VAE) allows agents to form perceptual categories and exchange signs within multimodal contexts. Moreover, fine-tuning the weight ratio to favor a modality that the model could learn and categorize more readily improved communication. Our evaluation of three multimodal approaches - mixture-of-experts (MoE), product-of-experts (PoE), and mixture-of-product-of-experts (MoPoE)-suggests an impact on the creation of latent spaces, the internal representations of agents. Our results from experiments with the MNIST + SVHN and Multimodal165 datasets indicate that combining the Gaussian mixture model (GMM), PoE multimodal VAE, and MH naming game substantially improved information sharing, knowledge formation, and data reconstruction.

Unsupervised generation of fashion editorials using deep generative model

This research intended to establish a new fashion-related artificial intelligence research topic concerning fashion editorials which could induce streams of further studies. A new fashion editorial dataset, which is a prerequisite in training an AI model, has been established in this study to meet the research purpose. A total of over 150K fashion editorials were initially collected and processed to satisfy necessary dataset conditions. A novel dataset of fashion editorials consisting of approximately 60K editorials is proposed through the process. In order to prove the adequacy of the new dataset, data distribution was analyzed and a generative model was selected and trained to attest that new fashion editorials can be created with the proposed editorial dataset. The results generated by the trained model were qualitatively investigated. The model has shown to have learned various features that compose editorials with the dataset, successfully generating fashion editorials. Quantitative evaluation with FID scores was conducted to support the selection of the generative model used for the qualitative assessment.

Evaluation of pseudo-healthy image reconstruction for anomaly detection with deep generative models: Application to brain FDG PET

Over the past years, pseudo-healthy reconstruction for unsupervised anomaly detection has gained in popularity. This approach has the great advantage of not requiring tedious pixel-wise data annotation and offers possibility to generalize to any kind of anomalies, including that corresponding to rare diseases. By training a deep generative model with only images from healthy subjects, the model will learn to reconstruct pseudo-healthy images. This pseudo-healthy reconstruction is then compared to the input to detect and localize anomalies. The evaluation of such methods often relies on a ground truth lesion mask that is available for test data, which may not exist depending on the application.<br>We propose an evaluation procedure based on the simulation of realistic abnormal images to validate pseudo-healthy reconstruction methods when no ground truth is available. This allows us to extensively test generative models on different kinds of anomalies and measuring their performance using the pair of normal and abnormal images corresponding to the same subject. It can be used as a preliminary automatic step to validate the capacity of a generative model to reconstruct pseudo-healthy images, before a more advanced validation step that would require clinician’s expertise. We apply this framework to the reconstruction of 3D brain FDG PET using a convolutional variational autoencoder with the aim to detect as early as possible the neurodegeneration markers that are specific to dementia such as Alzheimer’s disease.

DiffDec: Structure-Aware Scaffold Decoration with an End-to-End Diffusion Model.

In molecular optimization, one popular way is R-group decoration on molecular scaffolds, and many efforts have been made to generate R-groups based on deep generative models. However, these methods mostly use information on known binding ligands, without fully utilizing target structure information. In this study, we proposed a new method, DiffDec, to involve 3D pocket constraints by a modified diffusion technique for optimizing molecules through molecular scaffold decoration. For end-to-end generation of R-groups with different sizes, we designed a novel fake atom mechanism. DiffDec was shown to be able to generate structure-aware R-groups with realistic geometric substructures by the analysis of bond angles and dihedral angles and simultaneously generate multiple R-groups for one scaffold on different growth anchors. The growth anchors could be provided by users or automatically determined by our model. DiffDec achieved R-group recovery rates of 69.67% and 45.34% in the single and multiple R-group decoration tasks, respectively, and these values were significantly higher than competing methods (37.33% and 26.85%). According to the molecular docking study, our decorated molecules obtained a better average binding affinity than baseline methods. The docking pose analysis revealed that DiffDec could decorate scaffolds with R-groups that exhibited improved binding affinities and more favorable interactions with the pocket. These results demonstrated the potential and applicability of DiffDec in real-world scaffold decoration for molecular optimization.

Local Scaffold Diversity-Contributed Generator for Discovering Potential NLRP3 Inhibitors.

Deep generative models have become crucial tools in de novo drug design. In current models for multiobjective optimization in molecular generation, the scaffold diversity is limited when multiple constraints are introduced. To enhance scaffold diversity, we herein propose a local scaffold diversity-contributed generator (LSDC), which can be utilized to generate diverse lead compounds capable of satisfying multiple constraints. Compared to the state-of-the-art methods, molecules generated by LSDC exhibit greater diversity when applied to the generation of inhibitors targeting the NOD-like receptor (NLR) family, pyrin domain-containing protein 3 (NLRP3). We present 12 molecules, some of which feature previously unreported scaffolds, and demonstrate their reasonable docking binding modes. Consequently, the modification of selected scaffolds and subsequent bioactivity evaluation lead to the discovery of two potent NLRP3 inhibitors, A22 and A14, with IC50 values of 38.1 nM and 44.43 nM, respectively. And the oral bioavailability of compound A14 is very high (F is 83.09% in mice). This work contributes to the discovery of novel NLRP3 inhibitors and provides a reference for integrating AI-based generation with wet experiments.