Search

The nervous system, especially the human brain, is characterized by its highly complex network topology. The neurodevelopment of some of its features has been described in terms of dynamic optimization rules. We discuss the principle of adaptive rewiring, i.e., the dynamic reorganization of a network according to the intensity of internal signal communication as measured by synchronization or diffusion, and its recent generalization for applications in directed networks. These have extended the principle of adaptive rewiring from highly oversimplified networks to more neurally plausible ones. Adaptive rewiring captures all the key features of the complex brain topology: it transforms initially random or regular networks into networks with a modular small-world structure and a rich-club core. This effect is specific in the sense that it can be tailored to computational needs, robust in the sense that it does not depend on a critical regime, and flexible in the sense that parametric variation generates a range of variant network configurations. Extreme variant networks can be associated at macroscopic level with disorders such as schizophrenia, autism, and dyslexia, and suggest a relationship between dyslexia and creativity. Adaptive rewiring cooperates with network growth and interacts constructively with spatial organization principles in the formation of topographically distinct modules and structures such as ganglia and chains. At the mesoscopic level, adaptive rewiring enables the development of functional architectures, such as convergent-divergent units, and sheds light on the early development of divergence and convergence in, for example, the visual system. Finally, we discuss future prospects for the principle of adaptive rewiring. [ABSTRACT FROM AUTHOR]

Traditional antiviral peptide (AVP) discovery is a time-consuming and expensive process. This study introduces AVP-GPT, a novel deep learning method utilizing transformer-based language models and multimodal architectures specifically designed for AVP design. AVP-GPT demonstrated exceptional efficiency, generating 10,000 unique peptides and identifying potential AVPs within two days on a GPU system. Pre-trained on a respiratory syncytial virus (RSV) dataset, AVP-GPT successfully adapted to influenza A virus (INFVA) and other respiratory viruses. Compared to state-of-the-art models like LSTM and SVM, AVP-GPT achieved significantly lower perplexity (2.09 vs. 16.13) and higher AUC (0.90 vs. 0.82), indicating superior peptide sequence prediction and AVP classification. AVP-GPT generated a diverse set of peptides with excellent novelty and identified candidates with remarkably higher antiviral success rates than conventional design methods. Notably, AVP-GPT generated novel peptides against RSV and INFVA with exceptional potency, including four peptides exhibiting EC50 values around 0.02 uM—the strongest anti-RSV activity reported to date. These findings highlight AVP-GPT's potential to revolutionize AVP discovery and development, accelerating the creation of novel antiviral drugs. Future studies could explore the application of AVP-GPT to other viral targets and investigate alternative AVP design strategies. [ABSTRACT FROM AUTHOR]

Anomaly detection is one of the many challenging areas in cybersecurity. The anomaly can occur in many forms, such as fraudulent credit card transactions, network intrusions, and anomalous imageries or documents. One of the most common challenges in anomaly detection is the obscurity of the normal state and the lack of anomalous samples. Traditionally, this problem is tackled by using resampling techniques or choosing models that approximate the distribution of the normal states. Variational AutoEncoder (VAE) has been studied in anomaly detections despite being more suitable in generative tasks. This study aims to explore the usage of VAE in credit card anomaly detection and evaluate latent space sampling techniques. In this study, we evaluate the usage of the convolutional network-based VAE model on a credit card transaction dataset. We train two VAE models, one with a large number of normal data and one with a small number of anomalous data. We compare the performance of both VAE models and evaluate the latent space of both VAE models by rescaling them with reconstruction error vectors. We also compare the effectiveness of the VAE model with other anomaly detection models when they are trained on imbalanced dataset.

Proteins mediate their functions through chemical interactions; modeling these interactions, which are typically through sidechains, is an important need in protein design. However, constructing an all-atom generative model requires an appropriate scheme for managing the jointly continuous and discrete nature of proteins encoded in the structure and sequence. We describe an all-atom diffusion model of protein structure, Protpardelle, which represents all sidechain states at once as a "superposition" state; superpositions defining a protein are collapsed into individual residue types and conformations during sample generation. When combined with sequence design methods, our model is able to codesign all-atom protein structure and sequence. Generated proteins are of good quality under the typical quality, diversity, and novelty metrics, and sidechains reproduce the chemical features and behavior of natural proteins. Finally, we explore the potential of our model to conduct all-atom protein design and scaffold functional motifs in a backbone- and rotamer-free way. [ABSTRACT FROM AUTHOR]

The circle of Willis (CoW) is a network of cerebral arteries with significant inter-individual anatomical variations. Deep learning has been used to characterize and quantify the status of the CoW in various applications for the diagnosis and treatment of cerebrovascular disease. In medical imaging, the performance of deep learning models is limited by the diversity and size of training datasets. To address medical data scarcity, generative AI models have been applied to generate synthetic vessel neuroimaging data. However, the proposed methods produce synthetic data with limited anatomical fidelity or downstream utility in tasks concerning vessel characteristics.We adapted the StyleGANv2 architecture to 3D to synthesize Time-of-Flight Magnetic Resonance Angiography (TOF MRA) volumes of the CoW. For generative modeling, we used 1782 individual TOF MRA scans from 6 open source datasets. To train the adapted 3D StyleGAN model with limited data we employed differentiable data augmentations, used mixed precision and a cropped region of interest of size 32 × 128 × 128 to tackle computational constraints. The performance was evaluated quantitatively using the Fréchet Inception Distance (FID), MedicalNet distance (MD) and Area Under the Curve of the Precision and Recall Curve for Distributions (AUC-PRD). Qualitative analysis was performed via a visual Turing test. We demonstrated the utility of generated data in a downstream task of multiclass semantic segmentation of CoW arteries. Vessel segmentation performance was assessed quantitatively using the Dice coefficient and the Hausdorff distance.The best-performing 3D StyleGANv2 architecture generated high-quality and diverse synthetic TOF MRA volumes (FID: 12.17, MD: 0.00078, AUC-PRD: 0.9610). Multiclass vessel segmentation models trained on synthetic data alone achieved comparable performance to models trained using real data in most arteries. The addition of synthetic data to a baseline training set improved segmentation performance in underrepresented artery segments, similar to the addition of real data.In conclusion, generative modeling of the Circle of Willis via synthesis of 3D TOF MRA data paves the way for generalizable deep learning applications in cerebrovascular disease. In the future, the extensions of the provided methodology to other medical imaging problems or modalities with the inclusion of pathological datasets has the potential to advance the development of more robust AI models for clinical applications.

The nervous system, especially the human brain, is characterized by its highly complex network topology. The neurodevelopment of some of its features has been described in terms of dynamic optimization rules. We discuss the principle of adaptive rewiring, i.e., the dynamic reorganization of a network according to the intensity of internal signal communication as measured by synchronization or diffusion, and its recent generalization for applications in directed networks. These have extended the principle of adaptive rewiring from highly oversimplified networks to more neurally plausible ones. Adaptive rewiring captures all the key features of the complex brain topology: it transforms initially random or regular networks into networks with a modular small-world structure and a rich-club core. This effect is specific in the sense that it can be tailored to computational needs, robust in the sense that it does not depend on a critical regime, and flexible in the sense that parametric variation generates a range of variant network configurations. Extreme variant networks can be associated at macroscopic level with disorders such as schizophrenia, autism, and dyslexia, and suggest a relationship between dyslexia and creativity. Adaptive rewiring cooperates with network growth and interacts constructively with spatial organization principles in the formation of topographically distinct modules and structures such as ganglia and chains. At the mesoscopic level, adaptive rewiring enables the development of functional architectures, such as convergent-divergent units, and sheds light on the early development of divergence and convergence in, for example, the visual system. Finally, we discuss future prospects for the principle of adaptive rewiring.

Recent advances in multiplexed single-cell transcriptomics experiments facilitate the high-throughput study of drug and genetic perturbations. However, an exhaustive exploration of the combinatorial perturbation space is experimentally unfeasible. Therefore, computational methods are needed to predict, interpret, and prioritize perturbations. Here, we present the compositional perturbation autoencoder (CPA), which combines the interpretability of linear models with the flexibility of deep-learning approaches for single-cell response modeling. CPA learns to in silico predict transcriptional perturbation response at the single-cell level for unseen dosages, cell types, time points, and species. Using newly generated single-cell drug combination data, we validate that CPA can predict unseen drug combinations while outperforming baseline models. Additionally, the architecture's modularity enables incorporating the chemical representation of the drugs, allowing the prediction of cellular response to completely unseen drugs. Furthermore, CPA is also applicable to genetic combinatorial screens. We demonstrate this by imputing in silico 5,329 missing combinations (97.6% of all possibilities) in a single-cell Perturb-seq experiment with diverse genetic interactions. We envision CPA will facilitate efficient experimental design and hypothesis generation by enabling in silico response prediction at the single-cell level and thus accelerate therapeutic applications using single-cell technologies.

A multivariate distribution can be described by a triangular transport map from the target distribution to a simple reference distribution. We propose Bayesian nonparametric inference on the transport map by modeling its components using Gaussian processes. This enables regularization and uncertainty quantification of the map estimation, while resulting in a closed-form and invertible posterior map. We then focus on inferring the distribution of a nonstationary spatial field from a small number of replicates. We develop specific transport-map priors that are highly flexible and are motivated by the behavior of a large class of stochastic processes. Our approach is scalable to high-dimensional distributions due to data-dependent sparsity and parallel computations. We also discuss extensions, including Dirichlet process mixtures for flexible marginals. We present numerical results to demonstrate the accuracy, scalability, and usefulness of our methods, including statistical emulation of non-Gaussian climate-model output. for this article are available online. [ABSTRACT FROM AUTHOR]

The optimal structural design is imperative in order to minimize material consumption and reduce the environmental impacts of construction. Given the complexity in the formulation of structural design problems, the process of optimization is commonly performed using artificial intelligence (AI) global optimization, such as the genetic algorithm (GA). However, the integration of AI-based optimization, together with visual programming (VP), in building information modeling (BIM) projects warrants further investigation. This study proposes a workflow by combining structure analysis, VP, BIM, and GA to optimize trusses. The methodology encompasses several steps, including the following: (i) generation of parametric trusses in Dynamo VP; (ii) performing finite element modeling (FEM) using Robot Structural Analysis (RSA); (iii) retrieving and evaluating the FEM results interchangeably between Dynamo and RSA; (iv) finding the best solution using GA; and (v) importing the optimized model into Revit, enabling the user to perform simulations and engineering analysis, such as life cycle assessment (LCA) and quantity surveying. This methodology provides a new interoperable framework with minimal interference with existing supply-chain processes, and it will be flexible to technology literacy and allow architectural, engineering and construction (AEC) professionals to employ VP, global optimization, and FEM in BIM-based projects by leveraging open-sourced software and tools, together with commonly used design software. The feasibility of the proposed workflow was tested on benchmark problems and compared with the open literature. The outcomes of this study offer insight into the opportunities and limitations of combining VP, GA, FEA, and BIM for structural optimization applications, particularly to enhance structural efficiency and sustainability in construction. Despite the success of this study in developing a workable, user-friendly, and interoperable framework for the utilization of VP, GA, FEM, and BIM for structural optimization, the results obtained could be improved by (i) increasing the callback function speed between Dynamo and RSA through specialized application programming interface (API); and (ii) fine-tuning the GA parameters or utilizing other advanced global optimization and supervised learning techniques for the optimization. [ABSTRACT FROM AUTHOR]

Modeling and simulating the growth of the branching of tree species remains a challenge. With existing approaches, we can reconstruct or rebuild the branching architectures of real tree species, but the simulation of the growth process remains unresolved. First, we present a tree growth model to generate branching architectures that resemble real tree species. Secondly, we use a quantitative morphometric approach to infer the shape similarity of the generated simulations and real tree species. Within a functional–structural plant model, we implement a set of biological parameters that affect the branching architecture of trees. By modifying the parameter values, we aim to generate basic shapes of spruce, pine, oak and poplar. Tree shapes are compared using geometric morphometrics of landmarks that capture crown and stem outline shapes. Five biological parameters, namely xylem flow, shedding rate, proprioception, gravitysense and lightsense, most influenced the generated tree branching patterns. Adjusting these five parameters resulted in the different tree shapes of spruce, pine, oak, and poplar. The largest effect was attributed to gravity, as phenotypic responses to this effect resulted in different growth directions of gymnosperm and angiosperm branching architectures. Since we were able to obtain branching architectures that resemble real tree species by adjusting only a few biological parameters, our model is extendable to other tree species. Furthermore, the model will also allow the simulation of structural tree–environment interactions. Our simplifying approach to shape comparison between tree species, landmark geometric morphometrics, showed that even the crown–trunk outlines capture species differences based on their contrasting branching architectures. [ABSTRACT FROM AUTHOR]

Learning visual models of object classes conventionally require hundreds or a large number of training samples. Conventional gradient-based approaches for target recognition require lot of data to be trained on and require exhaustive training with high computational expense. Hence, when a new condition or untrained data is encountered, such systems inadequately misconfigure newly learned feature sets in the trained model. This misconfigures the structure of re-learned features and is then carried out in subsequent recognition stages. Thus, a development in this scenario with low training time will allow us to fend of this disadvantage. This study presents a new automatic target recognition framework that gives the enhanced performance of target-recognition system when several imaging sensors are connected with one another. This is in contrast with traditional automatic target recognition frameworks, which utilizes one-on-one computational model over synthetic-aperture radar image-processing systems. The work comprises of a learning-based classifications strategy when dealing with sharing of learned parameters over the network to discern critical changes in target-recognition performance by utilizing a novel one-shot learning-based reconfigurable learning network for image processing platform. This upgrades the networked connected CCTV and multiview synthetic-aperture radar image object identification and recognition process. [ABSTRACT FROM AUTHOR]

The simulation and optimization of metamaterial, known for their engineered properties in various applications, encounter intricate challenges due to complex microstructure interactions, extensive design spaces, and substantial computational requirements. To address these challenges, our research introduces a novel data-driven framework utilizing deep generative modeling to enhance the design process of metamaterials.Focusing on composite metamaterial with double negative coefficients of thermal expansion and Poisson's ratio, we apply an Alternative Active Phase and Objective functions (AAPO) method for deciding the initial metamaterial dataset. To enhance dataset diversity, a distortion filter is applied, broadening the range of design possibilities. Subsequently, we utilize a Variational Autoencoder (VAE), integrated with a regressor, to train on this diversified database. This training effectively maps complex unit cell geometries to a coherent latent space, simultaneously correlating them with continuous material properties.Our approach demonstrates robustness in multi-phase and multi-physics optimization as well as efficiency in generating specialized databases of unit cells. This framework is pivotal in systematically designing unit cells and multiscale systems, specifically aiming for distinct thermo-mechanical behavior targets.To mitigate the computational demands encountered during multiple design meta-materials via gradient-based topology optimization, we have integrated high-performance methods and automatic differentiation. This integration marks a significant advancement in the data-driven design of metamaterials, offering substantial practical and theoretical benefits in the field.

Artificial intelligence (AI) is increasingly applied across all disciplines of medicine, including dentistry. Oral health research is experiencing a rapidly increasing use of machine learning (ML), the branch of AI that identifies inherent patterns in data similarly to how humans learn. In contemporary clinical dentistry, ML supports computer-aided diagnostics, risk stratification, individual risk prediction, and decision support to ultimately improve clinical oral health care efficiency, outcomes, and reduce disparities. Further, ML is progressively used in dental and oral health research, from basic and translational science to clinical investigations. With an ML perspective, this review provides a comprehensive overview of how dental medicine leverages AI for diagnostic, prognostic, and generative tasks. The spectrum of available data modalities in dentistry and their compatibility with various methods of applied AI are presented. Finally, current challenges and limitations as well as future possibilities and considerations for AI application in dental medicine are summarized.

Vision-based grasping of unknown objects in unstructured environments is a key challenge for autonomous robotic manipulation. A practical grasp synthesis system is required to generate a diverse set of 6-DoF grasps from which a task-relevant grasp can be executed. Although generative models are suitable for learning such complex data distributions, existing models have limitations in grasp quality, long training times, and a lack of flexibility for task-specific generation. In this work, we present GraspLDM, a modular generative framework for 6-DoF grasp synthesis that uses diffusion models as priors in the latent space of a VAE. GraspLDM learns a generative model of object-centric $SE(3)$ grasp poses conditioned on point clouds. GraspLDM’s architecture enables us to train task-specific models efficiently by only re-training a small denoising network in the low-dimensional latent space, as opposed to existing models that need expensive re-training. Our framework provides robust and scalable models on both full and partial point clouds. GraspLDM models trained with simulation data transfer well to the real world without any further fine-tuning. Our models provide an 80% success rate for 80 grasp attempts of diverse test objects across two real-world robotic setups.

We propose a new methodology for the estimation of the rate distortion function (RDF), considering both continuous and discrete reconstruction spaces. The approach is input-space agnostic and does not require prior knowledge of the source distribution, nor the distortion function. Thus, our method is a general solution to the RDF estimation problem, while existing works focus on a specific domain. The approach leverages neural estimation and constrained optimization of mutual information to optimize a generative model of the input distribution. In continuous spaces we learn a sample generating model, while a probability mass function model is proposed for discrete spaces. Formal guarantees of the proposed method are explored and implementation details are discussed. We demonstrate our method’s superior performance on both high dimensional and large alphabet synthetic data. In contrast to existing works, our estimator readily adapts to the rate distortion perception framework, which is central to contemporary compression tasks. Consequently, our method strengthens the connection between information theory and machine learning, proposing new solutions to the problem of lossy compression.

Transcranial magnetic stimulation (TMS) is often applied to the motor cortex to stimulate a collection of motor evoked potentials (MEPs) in groups of peripheral muscles. The causal interface between TMS and MEP is the selective activation of neurons in the motor cortex; moving around the TMS ‘spot’ over the motor cortex causes different MEP responses. A question of interest is whether a collection of MEP responses can be used to identify the stimulated locations on the cortex, which could potentially be used to then place the TMS coil to produce chosen sets of MEPs. In this work we leverage our previous report on a 3D convolutional neural network (CNN) architecture that predicted MEPs from the induced electric field, to tackle an inverse imaging task in which we start with the MEPs and estimate the stimulated regions on the motor cortex. We present and evaluate five different inverse imaging CNN architectures, both conventional and generative, in terms of several measures of reconstruction accuracy. We found that one architecture, which we propose as M2M-InvNet, consistently achieved the best performance.

In the field of face frontalization, the model obtained by training on a particular dataset often underperforms on other datasets. This paper presents the Pre-trained Feature Transformation GAN (PFT-GAN), which is designed to fully utilize diverse facial feature information available from pre-trained face recognition networks. For that purpose, we propose the use of the feature attention transformation (FAT) module that effectively transfers the low-level facial features to the facial generator. On the other hand, in the hope of reducing the pre-trained encoder dependency, we attempt a new FAT module organization that accommodates the features from all pre-trained face recognition networks employed. This paper attempts evaluating the proposed work using the “independent critic” as well as “dependent critic”, which enables objective judgments. Experimental results show that the proposed method significantly improves the face frontalization performance and helps overcome the bias associated with each pre-trained face recognition network employed.

Traditional antiviral peptide (AVP) discovery is a time-consuming and expensive process. This study introduces AVP-GPT, a novel deep learning method utilizing transformer-based language models and multimodal architectures specifically designed for AVP design. AVP-GPT demonstrated exceptional efficiency, generating 10,000 unique peptides and identifying potential AVPs within two days on a GPU system. Pre-trained on a respiratory syncytial virus (RSV) dataset, AVP-GPT successfully adapted to influenza A virus (INFVA) and other respiratory viruses. Compared to state-of-the-art models like LSTM and SVM, AVP-GPT achieved significantly lower perplexity (2.09 vs. 16.13) and higher AUC (0.90 vs. 0.82), indicating superior peptide sequence prediction and AVP classification. AVP-GPT generated a diverse set of peptides with excellent novelty and identified candidates with remarkably higher antiviral success rates than conventional design methods. Notably, AVP-GPT generated novel peptides against RSV and INFVA with exceptional potency, including four peptides exhibiting EC50 values around 0.02 uM—the strongest anti-RSV activity reported to date. These findings highlight AVP-GPT’s potential to revolutionize AVP discovery and development, accelerating the creation of novel antiviral drugs. Future studies could explore the application of AVP-GPT to other viral targets and investigate alternative AVP design strategies.

The neural radiance field (NeRF) has shown promising results in preserving the fine details of objects and scenes. However, unlike explicit shape representations e.g., mesh, it remains an open problem to build dense correspondences across different NeRFs of the same category, which is essential in many downstream tasks. The main difficulties of this problem lie in the implicit nature of NeRF and the lack of ground-truth correspondence annotations. In this paper, we show it is possible to bypass these challenges by leveraging the rich semantics and structural priors encapsulated in a pre-trained NeRF-based GAN. Specifically, we exploit such priors from three aspects, namely (1) a dual deformation field that takes latent codes as global structural indicators, (2) a learning objective that regards generator features as geometric-aware local descriptors, and (3) a source of infinite object-specific NeRF samples. Our experiments demonstrate that such priors lead to 3D dense correspondence that is accurate, smooth, and robust. We also show that established dense correspondence across NeRFs can effectively enable many NeRF-based downstream applications such as texture transfer. [ABSTRACT FROM AUTHOR]

Person re-identification is very important for monitoring and tracking crowd movement to provide public security. However, re-identification in the presence of occlusion is a challenging area that has not received significant attention yet. In this work, we propose a plausible solution to this problem by developing effective techniques for occlusion detection and reconstruction from RGB images/videos using Deep Neural Networks. Specifically, a CNN-based occlusion detection model is used to detect the occluded frames in an input sequence, following which a Conv-LSTM model or an Autoencoder is employed to reconstruct the pixels corresponding to the occluded regions depending on whether the input frames are sequential or non-sequential. The quality of the reconstructed RGB frames is further refined using a DCGAN. Our method has been evaluated using four public data sets for cumulative rank-based accuracy and Dice score, and the qualitative reconstruction results are indeed appealing. Quantitative evaluation in terms of re-identification accuracy using a Siamese classifier shows a Rank-1 accuracy of over 70% after reconstructing the occlusion present in each of these datasets. A comparative study with popular state-of-the-art approaches also demonstrates the effectiveness of our work for use in real-life surveillance sites. [ABSTRACT FROM AUTHOR]

Recent rapid expansion of make‐on‐demand, purchasable, chemical libraries comprising dozens of billions or even trillions of molecules has challenged the efficient application of traditional structure‐based virtual screening methods that rely on molecular docking. We present a novel computational methodology termed HIDDEN GEM (HIt Discovery using Docking ENriched by GEnerative Modeling) that greatly accelerates virtual screening. This workflow uniquely integrates machine learning, generative chemistry, massive chemical similarity searching and molecular docking of small, selected libraries in the beginning and the end of the workflow. For each target, HIDDEN GEM nominates a small number of top‐scoring virtual hits prioritized from ultra‐large chemical libraries. We have benchmarked HIDDEN GEM by conducting virtual screening campaigns for 16 diverse protein targets using Enamine REAL Space library comprising 37 billion molecules. We show that HIDDEN GEM yields the highest enrichment factors as compared to state of the art accelerated virtual screening methods, while requiring the least computational resources. HIDDEN GEM can be executed with any docking software and employed by users with limited computational resources. [ABSTRACT FROM AUTHOR]

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.

Out-of-distribution (OOD) detection is crucial for enhancing the reliability of machine learning models when confronted with data that differ from their training distribution. In the image domain, we hypothesize that images inhabit manifolds defined by latent properties such as color, position, and shape. Leveraging this intuition, we propose a novel approach to OOD detection using a diffusion model to discern images that deviate from the in-domain distribution. Our method involves training a diffusion model using in-domain images. At inference time, we lift an image from its original manifold using a masking process, and then apply a diffusion model to map it towards the in-domain manifold. We measure the distance between the original and mapped images, and identify those with a large distance as OOD. Our experiments encompass comprehensive evaluation across various datasets characterized by differences in color, semantics, and resolution. Our method demonstrates strong and consistent performance in detecting OOD images across the tested datasets, highlighting its effectiveness in handling images with diverse characteristics. Additionally, ablation studies confirm the significant contribution of each component in our framework to the overall performance.

Antibody leads must fulfill multiple desirable properties to be clinical candidates. Primarily due to the low throughput in the experimental procedure, the need for such multi-property optimization causes the bottleneck in preclinical antibody discovery and development, because addressing one issue usually causes another. We developed a reinforcement learning (RL) method, named AB-Gen, for antibody library design using a generative pre-trained transformer (GPT) as the policy network of the RL agent. We showed that this model can learn the antibody space of heavy chain complementarity determining region 3 (CDRH3) and generate sequences with similar property distributions. Besides, when using human epidermal growth factor receptor-2 (HER2) as the target, the agent model of AB-Gen was able to generate novel CDRH3 sequences that fulfill multi-property constraints. Totally, 509 generated sequences were able to pass all property filters, and three highly conserved residues were identified. The importance of these residues was further demonstrated by molecular dynamics simulations, consolidating that the agent model was capable of grasping important information in this complex optimization task. Overall, the AB-Gen method is able to design novel antibody sequences with an improved success rate than the traditional propose-then-filter approach. It has the potential to be used in practical antibody design, thus empowering the antibody discovery and development process. The source code of AB-Gen is freely available at Zenodo (https://doi.org/10.5281/zenodo.7657016) and BioCode (https://ngdc.cncb.ac.cn/biocode/tools/BT007341).