Single Vector Research Articles

The Detection of Kiwifruit Sunscald Using Spectral Reflectance Data Combined with Machine Learning and CNNs

Sunscald in kiwifruit, an environmental stress caused by solar radiation during the summer, reduces fruit quality and yields and causes economic losses. The efficient and timely detection of sunscald and similar diseases is a challenging task but helps to implement measures to control stress. This study provides high-precision detection models and relevant spectral information on kiwifruit physiology for similar statuses, including early-stage sunscald, late-stage sunscald, anthracnose, and healthy. Primarily, in the laboratory, 429 groups of spectral reflectance data for leaves of four statuses were collected and analyzed using a hyperspectral reflection acquisition system. Then, multiple modeling approaches, including combined preprocessing methods, feature extraction algorithms, and classification algorithms, were designed to extract bands and evaluate the performance of the models to detect the statuses of kiwifruit. Finally, the detection of different stages of kiwifruit sunscald under anthracnose interference was accomplished. As influential bands, 694–713 nm, 758–777 nm, 780–799 nm, and 1303–1322 nm were extracted. The overall accuracy, precision, recall, and F1-score values of the models reached 100%, demonstrating an ability to detect all statuses with 100% accuracy. It was concluded that the combined processing of moving average and standard normal variable transformations (MS) could significantly improve the data; the near-infrared support vector machine and visible convolutional neural network with MS (NIR-MS-SVM and VIS-MS-CNN) were established as high-precision detection techniques for the classification of similar kiwifruit statuses, demonstrating 25.58% higher accuracy than the single support vector machine. The VIS-MS-CNN model reached convergence with a stable cross-entropy loss of 0.75 in training and 0.77 in validation. The techniques developed in this study will improve orchard management efficiency and yields and increase researchers’ understanding of kiwifruit physiology.

A method for constructing word sense embeddings based on word sense induction

Polysemy is an inherent characteristic of natural language. In order to make it easier to distinguish between different senses of polysemous words, we propose a method for encoding multiple different senses of polysemous words using a single vector. The method first uses a two-layer bidirectional long short-term memory neural network and a self-attention mechanism to extract the contextual information of polysemous words. Then, a K-means algorithm, which is improved by optimizing the density peaks clustering algorithm based on cosine similarity, is applied to perform word sense induction on the contextual information of polysemous words. Finally, the method constructs the corresponding word sense embedded representations of the polysemous words. The results of the experiments demonstrate that the proposed method produces better word sense induction than Euclidean distance, Pearson correlation, and KL-divergence and more accurate word sense embeddings than mean shift, DBSCAN, spectral clustering, and agglomerative clustering.

Leverage of nuclease-deficient CasX for preventing pathological angiogenesis

Gene editing with a CRISPR/Cas system is a novel potential strategy for treating human diseases. Pharmacological inhibition of phosphoinositide 3-kinase (PI3K) δ suppresses retinal angiogenesis in a mouse model of oxygen-induced retinopathy. Here we show that an innovative system of adeno-associated virus (AAV)-mediated CRISPR/nuclease-deficient (d)CasX fused with the Krueppel-associated box (KRAB) domain is leveraged to block (81.2%± 6.5%) invitro expression of p110δ, the catalytic subunit of PI3Kδ, encoded by Pik3cd. This CRISPR/dCasX-KRAB (4, 269bp) system is small enough to be fit into a single AAV vector. We then document that recombinant AAV serotype (rAAV)1 efficiently transduces vascular endothelial cells from pathologic retinal vessels, which show high expression of p110δ; furthermore, we demonstrate that blockade of retinal p110δ expression by intravitreally injected rAAV1-CRISPR/dCasX-KRAB targeting the Pik3cd promoter prevents (32.1%± 5.3%) retinal p110δ expression as well as pathological retinal angiogenesis in a mouse model of oxygen-induced retinopathy. These data establish a strong foundation for treating pathological angiogenesis by AAV-mediated CRISPR interference with p110δ expression.

Integrated Controller for Divert Thruster Flight Propulsion Control Systems

Future generation missile calls for development of fast response propulsion based control systems to achieve hit-to-kill miss distance against highly manoeuvrable targets covering high-endo to exo-atmosphere regions. Control systems based on solid propulsion are developed connecting multiple pintle nozzles and control thrust vector is generated with distribution hot gasses in pitch/yaw directions. Differential distribution of hot gasses among multiple nozzles is complex due to time varying coupled non-linear chamber pressure dynamics with propellant combustion and nozzles’ flow phenomena and methods based on linear control may not ensure stability in wide operating conditions. There is need to develop "Unified Controller" based on non-linear formulation to ensure control stability in wide operating conditions of various mission applications, with advanced features for new generation missiles to counter various futuristic targets, and achieve hit-to-kill capabilities following integrated approach in order to improve the total system performance. Non-linear algorithms proposed in the recent literature do not exploit synergism between plant and actuation systems dynamics. Present paper presents novel non-linear feedback linearization technique, Dynamic Inversion, a relatively straightforward approach in Integrated Control Framework (ICF) implementing inherent time scale separation between solid motor plant and hot gas flow control valve actuation system. To ensure stability in various operating conditions and assure robust performance in the presence of parameter inaccuracies, controller is augmented with model-following neuro-adaptive design. Prototype systems: Thrust Controlled Motor (TCM) and Divert Thruster Flight Propulsion Control Systems (DTFPCS) are developed to demostrate/validate "Unified Controller" for missile control system. Closed loop control of chamber pressure is demonstrated with single pintle nozzle system and control thrust vector generation with multiple nozzle systems. Model update takes place adaptively online and unlike a typical indirect adaptive control approach, there is no need to update the individual parameters explicitly. Instead the inaccuracy in the system output dynamics is captured directly and then used in modifying the control. Based on present work, missilecontrol systems for various applications can be developed with limited controller changes and thereby minimizing number of developmental tests and cost.

Region-Aware Hierarchical Latent Feature Representation Learning-Guided Clustering for Hyperspectral Band Selection.

Hyperspectral band selection aims to identify an optimal subset of bands for hyperspectral images (HSIs). For most existing clustering-based band selection methods, they directly stretch each band into a single feature vector and employ the pixelwise features to address band redundancy. In this way, they do not take full consideration of the spatial information and deal with the importance of different regions in HSIs, which leads to a nonoptimal selection. To address these issues, a region-aware hierarchical latent feature representation learning-guided clustering (HLFC) method is proposed. Specifically, in order to fully preserve the spatial information of HSIs, the superpixel segmentation algorithm is adopted to segment HSIs into multiple regions first. For each segmented region, the similarity graph is constructed to reflect the bands-wise similarity, and its corresponding Laplacian matrix is generated for learning low-dimensional latent features in a hierarchical way. All latent features are then fused to form a unified feature representation of HSIs. Finally, k -means clustering is utilized on the unified feature representation matrix to generate multiple clusters from which the band with maximum information entropy is selected to form the final subset of bands. Extensive experimental results demonstrate that the proposed clustering method can achieve superior performance than the state-of-the-art representative methods on the band selection. The demo code of this work is publicly available at https://github.com/WangJun2023/HLFC.

Line spectrum extraction method of underwater targets based on the polarization characteristic

In order to suppress the environmental noise interference of underwater targets, a line spectrum extraction method is proposed based on the polarization characteristic of underwater acoustic signals received by a single vector sensor. In this paper, the polarization filtering function of a fourth-order cumulant diagonal slice is developed, and the fourth-order cumulant diagonal slice characteristic spectrum corresponding to the maximum singular value decomposition is used to extract the line spectrum. The simulation and experimental results show that the proposed method can extract the line spectrum signal adaptively, which has a good abatement effect on environmental noise interference.

Direction Estimation and Tracking of Coherent Sources Using a Single Acoustic Vector Sensor

A single acoustic vector sensor (AVS) cannot be used to find the direction-of-arrival (DOA) of two or more coherent (fully correlated) sources. We have proposed a technique for estimating DOAs (in 2D geometry) of two simultaneous coherent sources using single AVS under the assumption that acoustic sources enter in the field sequentially. The DOA estimation has been investigated with two different configurations of AVS, each consisting of three microphones in a plane. The technique has been also applied in tracking (a) an acoustic source in the presence of stationary interfering coherent source and (b) two coherent sources when the sources are changing their locations alternatively. The experimental environment has been generated using the Finite-Element Method tool viz. COMSOL to corroborate the proposed scheme.

Anomaly Prediction over Human Crowded Scenes via Associate-Based Data Mining and K-Ary Tree Hashing

Anomaly detection and behavioral recognition are key research areas widely used to improve human safety. However, in recent times, with the extensive use of surveillance systems and the substantial increase in the volume of recorded scenes, the conventional analysis of categorizing anomalous events has proven to be a difficult task. As a result, machine learning researchers require a smart surveillance system to detect anomalies. This research introduces a robust system for predicting pedestrian anomalies. First, we acquired the crowd data as input from two benchmark datasets (including Avenue and ADOC). Then, different denoising techniques (such as frame conversion, background subtraction, and RGB-to-binary image conversion) for unfiltered data are carried out. Second, texton segmentation is performed to identify human subjects from acquired denoised data. Third, we used Gaussian smoothing and crowd clustering to analyze the multiple subjects from the acquired data for further estimations. The next step is to perform feature extraction to multiple abstract cues from the data. These bag of features include periodic motion, shape autocorrelation, and motion direction flow. Then, the abstracted features are mapped into a single vector in order to apply data optimization and mining techniques. Next, we apply the associate-based mining approach for optimized feature selection. Finally, the resultant vector is served to the k-ary tree hashing classifier to track normal and abnormal activities in pedestrian crowded scenes.

Inception-based global context attention network for the classification of coffee leaf diseases

Coffee is a significant global commodity that is consumed in large quantities on a daily basis, and it ranks as the second-most important product in global trade. The leaves of coffee plants are vulnerable to fungal and pest attacks that can harm their photosynthetic regions, leading to the production of low-quality, diseased beans. Accurate diagnosis of coffee disease is crucial in determining suitable remedial action. Different pathogens require different pesticides or other forms of treatment, and a precise diagnosis can guarantee that the necessary steps are taken to prevent further damage to the plant. The major objective of this research is to develop a disease detection approach for coffee leaves based on deep learning techniques. The proposed network is developed with inception modules, a global context module, and a multi-head attention module to achieve precise classification. Through the simultaneous application of different filter sizes and pooling operations on the input, inception modules facilitate the proposed network in extracting features at multiple scales and acquiring significant feature maps at different levels of abstraction. The Global Context Block employs a channel attention mechanism to generate a single feature vector that modulates the input feature maps to obtain high-level contextual information. Finally, the multi-head attention module captures complex relationships between features from different subsets and aggregates them to form a more powerful representation of the input. The experimental results indicated that the proposed network, which was trained on the BRACOL dataset, outperformed existing networks in detecting coffee leaf disease. It achieved an accuracy of 98.57% and a F1 score of 98.55%.

Robust Key Update With Controllable Accuracy Using Support Vector Machine for Secure OFDMA-PON

In this paper, we first propose a novel secret key update scheme only with a single support vector machine (SVM). To achieve robust and efficient key management for secure orthogonal frequency division multiple access-based passive optical network (OFDMA-PON). The key sequences required to be shared between multiple optical network units (ONUs) and optical line terminal (OLT) are converted into partial quadrature amplitude modulation (QAM) symbol phases, and then masked into the whole upstream signal using two-dimensional chaos. The session keys can be distributed in real-time without consuming exclusive secure channels and resulting in data rate loss. A well-trained support vector machine (SVM) implemented at the receiver enables error-free key decision. Moreover, the numerous cross validations are conducted to verify the superior generalization performance under various parameters including key length and modulation format. The 16-QAM, 32-QAM, and 64-QAM OFDM signal transmissions are successfully demonstrated over 25 km standard single mode fiber (SSMF) by an experiment. The results show that the controllable restored accuracy of key can be realized by regulating the pivotal algorithm parameter to flexibly adapt to different system conditions, thus this scheme can be a promising efficient key update technology for future secure PON.

Assessment of Therapeutic Potential of a Dual AAV Approach for Duchenne Muscular Dystrophy.

Duchenne muscular dystrophy (DMD) is a yet incurable rare genetic disease that affects the skeletal and cardiac muscles, leading to progressive muscle wasting and premature death. DMD is caused by the lack of dystrophin, a muscle protein essential for the biochemical support and integrity of muscle fibers. Gene replacement strategies for Duchenne muscular dystrophy (DMD) employing the adeno-associated virus (AAV) face the challenge imposed by the limited packaging capacity of AAV, only allowing the accommodation of a short version of dystrophin (µDys) that is still far removed from correcting human disease. The need to develop strategies leading to the expression of a best performing dystrophin variant led to only few studies reporting on the use of dual vectors, but none reported on a method to assess in vivo transgene reconstitution efficiency, the degree of which directly affects the use of safe AAV dosing. We report here on the generation of a dual AAV vector approach for the expression of a larger dystrophin version (quasidystrophin) based on homologous recombination, and the development of a methodology employing a strategic droplet digital PCR design, to determine the recombination efficiency as well as the occurrence of unwanted concatemerization events or aberrant expression from the single vectors. We demonstrated that, upon systemic delivery in the dystrophic D2.B10-Dmdmdx/J (DBA2mdx) mice, our dual AAV approach led to high transgene reconstitution efficiency and negligible Inverted Terminal Repeats (ITR)-dependent concatemerization, with consequent remarkable protein restoration in muscles and improvement of muscle pathology. This evidence supports the suitability of our system for gene therapy application and the potential of this methodology to assess and improve the feasibility for therapeutic translation of multiple vector approaches.

Unifying Sentence Transformer Embedding and Softmax Voting Ensemble for Accurate News Category Prediction

The study focuses on news category prediction and investigates the performance of sentence embedding of four transformer models (BERT, RoBERTa, MPNet, and T5) and their variants as feature vectors when combined with Softmax and Random Forest using two accessible news datasets from Kaggle. The data are stratified into train and test sets to ensure equal representation of each category. Word embeddings are generated using transformer models, with the last hidden layer selected as the embedding. Mean pooling calculates a single vector representation called sentence embedding, capturing the overall meaning of the news article. The performance of Softmax and Random Forest, as well as the soft voting of both, is evaluated using evaluation measures such as accuracy, F1 score, precision, and recall. The study also contributes by evaluating the performance of Softmax and Random Forest individually. The macro-average F1 score is calculated to compare the performance of different transformer embeddings in the same experimental settings. The experiments reveal that MPNet versions v1 and v3 achieve the highest F1 score of 97.7% when combined with Random Forest, while T5 Large embedding achieves the highest F1 score of 98.2% when used with Softmax regression. MPNet v1 performs exceptionally well when used in the voting classifier, obtaining an impressive F1 score of 98.6%. In conclusion, the experiments validate the superiority of certain transformer models, such as MPNet v1, MPNet v3, and DistilRoBERTa, when used to calculate sentence embeddings within the Random Forest framework. The results also highlight the promising performance of T5 Large and RoBERTa Large in voting of Softmax regression and Random Forest. The voting classifier, employing transformer embeddings and ensemble learning techniques, consistently outperforms other baselines and individual algorithms. These findings emphasize the effectiveness of the voting classifier with transformer embeddings in achieving accurate and reliable predictions for news category classification tasks.

Paired contrastive feature for highly reliable offline signature verification

Signature verification requires high reliability. Especially in the writer-independent scenario with the skilled-forgery-only condition, achieving high reliability is challenging but very important. In this paper, we propose to apply two machine learning frameworks, learning with rejection and top-rank learning, to this task because they can suppress ambiguous results and thus give only reliable verification results. Since those frameworks accept a single input, we transform a pair of genuine and query signatures into a single feature vector, called Paired Contrastive Feature (PCF). PCF internally represents similarity (or discrepancy) between the two paired signatures; thus, reliable machine learning frameworks can make reliable decisions using PCF. Through experiments on three public signature datasets in the offline skilled-forgery-only writer-independent scenario, we evaluate and validate the effectiveness and reliability of the proposed models by comparing their performance with a state-of-the-art model.

Direction‐of‐arrival estimation for wideband linear frequency modulated signals with weighted l1‐norm in the fractional Fourier domain

AbstractDirection‐of‐arrival (DOA) estimation methods for wideband linear frequency modulated (LFM) signals based on fractional Fourier transform (FRFT) typically use the single snapshot vector in the fractional Fourier (FRF) domain. However, the low‐rank problem of the second‐order statistics in the single snapshot vector can decrease the DOA estimation resolution and cause issues in DOA estimation for coherent sources. In order to address these problems, a weighted l1‐norm method based on FRFT is proposed. The proposed method utilises the energy aggregation property of FRFT for LFM signals. This allows for obtaining a single snapshot vector in the FRF domain. Subsequently, a sparse representation DOA estimation model is established by taking advantage of the sparsity of the target distribution in the spatial domain. The DOA estimation is then obtained by minimising the l1‐norm. To further enhance the DOA estimation performance, we construct a matrix by rearranging the elements of the eigenvectors corresponding to the nonzero eigenvalue of the covariance matrix in the FRF domain. The rank of this matrix is equal to the number of sources, ensuring the orthogonality of the subspace obtained by the reconstructed matrix. Finally, a weighted matrix is constructed by exploiting the subspace orthogonality principle to adjust the sparsity penalty of the l1 constraint model. Simulation results demonstrate that our method is capable of achieving high angular resolution and accuracy even at low SNR and can process coherent signals directly.

High-resolution DOA estimation achieved by a single acoustic vector sensor under anisotropic noise

The acoustic vector sensor (AVS) has a wide application in the direction of arrival (DOA) estimation of sound sources including aircraft, underwater or on-land vehicles, search-and-rescue efforts, etc. The performance of only one acoustic vector sensor can compare favorably with a traditional acoustic array in DOA estimation. The anisotropic noise problem of the vector sensor has had a severe affective on the estimation accuracy and has caused little attention until now. In this paper, an AVS measurement system is designed, and an Anisotropic Sparse Bayesian Learning (SBL) method (the A-SBL method) is proposed to model and estimate the anisotropic noise of the AVS system. A high-resolution DOA estimation method is also introduced based on the noise estimation algorithm. The experimental results exhibit the necessity of modeling the anisotropic noise, and the estimation accuracy of the proposed A-SBL method based on the noise model is verified. The comparison of DOA methods shows the superiority of the proposed method, in which the DOA estimation error is not more than 1°. The proposed method’s utility may be found in applications for localization or tracking of targets with prominent acoustic characteristics when the environment is complicated.

A Novel Set-Valued Sensor Fault Diagnosis Method for Lithium-Ion Battery Packs in Electric Vehicles

Sensor fault diagnosis is of great significance to ensure safe battery operation. This paper proposes a novel sensor fault diagnosis method that achieves the simultaneous fault detection, fault source and type identification, and fault estimation in a comprehensive way. Specifically, a set-valued observer, featuring a state predictor and a state estimator, is first constructed and designed to guarantee the inclusion of the unavailable actual battery state due to unknown modeling errors and noises at every instant of time. Compared with the traditional observers, a distinct feature of the proposed one lies in that the calculated state predictions and estimations of the battery system at each time step are ellipsoidal sets in state space rather than single vectors. The boundedness of state prediction and estimation errors is formally proved, and the tractable design criteria for determining the real-time optimal prediction and estimation ellipsoids are also derived. As for diagnosis algorithm, fault detection is implemented based on the intersection between the prediction and estimation ellipsoids. Then, a two-layer Pearson correlation coefficient analysis mechanism is developed to identify the source and type of sensor faults. Another set-valued observer based on an augmented battery model is further designed to estimate the fault level. Finally, experimental studies of a battery cell under different sensor fault sources, types and values are elaborated to verify the effectiveness of the proposed method.

Bayes' Theorem for Multi-Bearing Faults Diagnosis

During the process of fault diagnosis for automated machinery, support vector machines is one of the suitable choices to categorize multiple faults for machinery. Regardless of the volume of sampling data, support vector machines can handle a high number of input features. It was learned that support vector machines could only sense binary fault classification (such as faulty or healthy). However, the classification accuracy was found to be lower when using support vector machines to diagnose multi-bearing faults classifications. This is because the multiple classification problem will be reduced into several sub-problems of binary classification when support vector machines adapt to multi-bearing faults classifications. From there, many contradictory results will occur from every support vector machine model. In order to solve the situation, the combination of Support Vector Machines and Bayes’ Theorem is introduced to every single support vector machine model to overcome the conflicting results. This method will also increase classification accuracy. The proposed Support Vector Machines - Bayes’ Theorem method has resulted in an increase in the accuracy of the fault diagnosis model. The analysis result has shown an accuracy from 72% to 95%. It proved that Support Vector Machines - Bayes’ Theorem continuously eliminates and refines conflicting results from the original support vector machine model. Compared to the existing support vector machine, the proposed Support Vector Machines - Bayes’ Theorem has proven its effectiveness in diagnosing the multi-bearing faults problem classification.

Targeted gene deletions in the dimorphic fungal pathogen Histoplasma using an optimized episomal CRISPR/Cas9 system.

The rapid development of CRISPR/CRISPR-associated (Cas) systems has revolutionized the ability to produce genetic mutations in a desired locus, particularly in organisms with low rates of homologous recombination. Histoplasma is an important respiratory and systemic fungal pathogen that has few reverse genetic options. We describe an optimized CRISPR/Cas system for the efficient generation of mutations in desired genes. The limited requirements for CRISPR/Cas, namely a gene-targeting guide RNA (gRNA) and expression of a Cas endonuclease, enabled both the gRNA and the Streptococcus pyogenes Cas9 gene to be expressed from a single episomal vector. The gRNAs are expressed from a strong Pol(II) promoter, a critical parameter for increasing the recovery of mutated genes, and processed into the mature gRNA by ribozymes in the mRNA. Expression of dual-tandem gRNAs facilitates the generation of gene deletions at a good frequency which can be detected by PCR-based screening of pooled isolates resulting in the isolation of marker-less deletion mutants. The CRISPR/Cas system is encoded on an episomal telomeric vector facilitating curing strains of the CRISPR/Cas vector upon generation of the mutant. We demonstrate the successful application of this CRISPR/Cas system in diverse Histoplasma species and applicable for multiple genes. The optimized system shows promise for accelerating reverse genetic studies in Histoplasma spp. IMPORTANCE The ability to eliminate gene product functions is central to understanding molecular mechanisms. In the fungal pathogen Histoplasma, methods to inactivate or deplete gene products are inefficient, which hampers progress in defining Histoplasma's virulence mechanisms. We describe an efficient CRISPR/Cas-based system for generating gene deletions in Histoplasma and show its validation on multiple genes with selectable and non-selectable phenotypes.

Novel comprehensive variable selection algorithm based on multi-weight vector optimal selection and bootstrapping soft shrinkage

The dimensionality of spectral data is increasing with the advancements in spectral technology. Therefore, there is an urgent need to develop high-performance variable selection algorithms for chemometrics applications. This study proposes a novel multi-weight optimal-bootstrap soft shrinkage (MWO-BOSS) method for variable selection based on the bootstrap soft shrinkage (BOSS) algorithm, comprising three effective improvement strategies. First, the optimal weight vector of six weight vectors are used as weights of the selection variables, rather than the absolute value of the regression coefficients based only on a single weight vector. Second, in each loop, a step-by-step strategy is implemented to determine the optimal set of variables. Finally, a smoothing operation is added to the weight vector to improve the anti-noise performance of the algorithm. The performance of the MWO-BOSS algorithm was tested on the four spectral datasets corn protein, corn oil, soil, and beer and compared with six high-performance algorithms, namely interval partial least squares (iPLS), Moving Window Partial Least-Squares(MWPLS), competitive adaptive reweighted sampling (CARS), variable combinatorial population analysis (VCPA), VCPA-IRIV and BOSS. The results show that the MWO-BOSS algorithm effectively improves the predictive ability of the model, with MWO-BOSS-Step-S providing the best results among the four tested datasets.

Translating Planar Heterocycles into Three-Dimensional Analogs by Photoinduced Hydrocarboxylation.

The rapid preparation of complex three-dimensional (3D) heterocyclic scaffolds is a key challenge in modern medicinal chemistry. Despite the increased probability of clinical success for small molecule therapeutic candidates with increased 3D complexity, new drug targets remain dominated by flat molecules due to the abundance of coupling reactions available for their construction. In principle, heteroarene hydrofunctionalization reactions offer an opportunity to transform readily accessible planar molecules into more three-dimensionally complex analogs through the introduction of a single molecular vector. Unfortunately, dearomative hydrofunctionalization reactions remain limited. Herein, we report a new strategy to enable the dearomative hydrocarboxylation of indoles and related heterocycles. This reaction represents a rare example of a heteroarene hydrofunctionalization that meets the numerous requirements for broad implementation in drug discovery. The transformation is highly chemoselective, broad in scope, operationally simple, and readily amenable to high-throughput experimentation (HTE). Accordingly, this process will allow existing libraries of heteroaromatic compounds to be translated into diverse 3D analogs and enable exploration of new classes of medicinally relevant molecules.