We thank MD Kazim Okan Dolu for their interest in our published work and for the opportunity to clarify several points regarding software versioning, benchmark independence, secondary-structure generation, binding-free-energy terminology, and negative-control design. We acknowledge that several descriptions in the original manuscript could be stated more precisely. These clarifications improve the transparency and reproducibility of the study, but do not affect the reported analyses or main conclusions.

Drug discovery is a time-consuming, expensive, and high-risk process. Recent advances in artificial intelligence (AI) have enabled major breakthroughs in small-molecule and protein therapeutics. However, AI-driven design of aptamer drugs remains largely unexplored. Aptamers are short (15-100 nt) single-stranded DNAs or RNAs that exhibit high binding affinity, high specificity, and low immunogenicity, making them promising candidates for disease (such as cancer) therapeutics. Compared with protein-ligand or protein-protein systems, protein-aptamer complexes are under-represented in public structural databases, and aptamers themselves are highly flexible and relatively large molecules. These characteristics present distinct challenges for AI-based structural modeling. Here, we systematically evaluate recent AI frameworks, including AlphaFold3, Chai-1, Boltz-2, and RoseTTAFold2NA, along with a template-based approach, in predicting protein-aptamer complex structures and estimating binding free energies. We establish an independent benchmark to assess their performance in structural accuracy, stability, and energetic consistency. This study provides a foundation for the application of AI in aptamer drug design and offers a reference framework for future research in nucleic-acid therapeutics and biomolecular modeling.

Lung nodules are the main lesions of the lung, and conditions of the lung can be directly displayed through CT images. As lung nodules are only more widely detectable once they have moved to other lung sections, it is highly challenging to anticipate the incidence of lung cancer at the beginning stages. In order to recognize lung nodules in a better and more timely manner and to prevent delays in the diagnosis of lung nodules, this paper proposes a new deep learning-based lung nodule detection model (MSF-YOLO). The model first integrates fusion of global and local features module (FGLM) into the backbone of YOLOv10n to improve the feature extraction capability of the model. Then, a new lateral cross-scale feature fusion structure (TiFPN) is added to the original feature pyramid to realize the learning of target features at different scales. Finally, the Inner Wise-MPDIoU loss function is proposed to improve the detection accuracy. The experimental results show that the proposed MSF-YOLO achieves [email protected] of 94.8% and [email protected]:0.95 of 62.9% (compared with YOLOv10-n increases by 4.6% and 4.5%, respectively)on the LIDC-IDRI dataset.

Recent advances in diffusion models have enhanced multimodal-guided visual generation, enabling customized subject insertion that seamlessly “brushes” user-specified objects into a given image guided by textual prompts. However, existing methods often struggle to insert customized subjects with high fidelity and align results with the user’s intent through textual prompts. In this work, we propose In-Context Brush, a zero-shot framework for customized subject insertion by reformulating the task within the paradigm of in-context learning. Without loss of generality, we formulate the object image and the textual prompts as cross-modal demonstrations, and the target image with the masked region as the query. The goal is to inpaint the target image with the subject aligning textual prompts without model tuning. Building upon a pretrained MMDiT-based inpainting network, we perform test-time enhancement via dual-level latent space manipulation: intra-head latent feature shifting within each attention head that dynamically shifts attention outputs to reflect the desired subject semantics and inter-head attention reweighting across different heads that amplifies prompt controllability through differential attention prioritization. Extensive experiments and applications demonstrate that our approach achieves superior identity preservation, text alignment, and image quality compared to existing state-of-the-art methods, without requiring dedicated training or additional data collection. Project page: https://yuci-gpt.github.io/In-Context-Brush/.

Sequential recommendation aims to predict the next item based on historical user interactions, which is crucial for online e-commerce platforms. Most existing methods rely on singular type of interactions for sequence modeling to understand user preference, overlooking the heterogeneous behavior information between users and items. From empirical analysis, we discovered that incorporating behavioral context into learning process of user preference is effective. However, oversimplified feature approaches are limited to model performance. To this end, we propose an Attention Mixture based Multi-scale Transformer framework to address this limitation. Specifically, we devise an attention mixture module that jointly considers user-item interactions and behavioral context to capture users’ personalized multi-behavior dependencies. It allows to perform effective behavior-aware sequence modeling. Then, we incorporate the attention mixture module into a multi-scale transformer to capture the periodic patterns in multi-behavior sequences. Empirical results on three real-world e-commerce datasets demonstrate the effectiveness of the proposed method.

Learning dynamic user preferences has become an important component for E-commerce platforms to make sequential recommendations. Many works have been focused on modeling users’ purchase behavior for sequential recommendations, while less work has considered users’ dynamic behavioral intentions. With this concern, we introduce a new recommendation problem, Behavior-Oriented Recommendation (BOR), where recommenders must retrieve items based on users’ diverse behavioral intentions. In this paper, We figure out that due to the discrepancy between BOR and traditional scenarios, existing recommendation models are struggling in discovering preference drift due to changes in behavioral intentions. To tackle this problem, we propose a Target-aware Feature Selection Network (TFSN). More specifically, a sequence encoder based on low-rank self-attention is designed to enhance the sequence feature representation. Then, the Decision Transformer architecture is employed to model the complex dependencies between sequence features and target behaviors. Furthermore, a weighted contrastive learning method is proposed to distinguish the preference representations under various targets. Extensive experiments on three real-world E-commerce datasets demonstrate the superiority of TFSN over state-of-the-art methods.

Multi-behavior sequential recommendation aims to predict users' next interested item, by learning dynamic user preferences within their multi-behavior interaction sequences. Users' dynamic preferences are decided by both stable long-term and variable short-term interests. Early efforts towards entangling these two aspects, which may lead to inferior recommendation accuracy and interpretability. Moreover, they ignore the differences in importance between different samples consequently limiting the model-fitting performance. With this concern, we propose a new recommendation framework Multi-_Behavior Interest Matching Network (MB-IMN) to overcome these limitations. Specifically, we model two interest aspects explicitly: a transformer with attention fusion encodes behavior-aware sequential patterns for short-term interest, and a graph learning paradigm is developed to capture multi-behavior interaction semantics for long-term interest. Furthermore, we devise a novel loss function that automatically determines sample importance based on a predefined modulating factor, thus reweighting the samples accordingly. Empirical results on three real-world e-commerce datasets demonstrate the effectiveness of proposed framework. Our implementation code is released at https://github.com/tripiggyo1/MB-IMN.

With the increasing maturity of smart grid and computer vision technologies, the use of mobile edge devices in collaboration with “cloud” to monitor the safety condition of construction workers wearing helmets has been widely used in power system construction scenarios. In this paper, based on YOLOv5, a new target detector-L-YOLOv5 is proposed to improve the accuracy and efficiency of the detector by optimizing key components. First, the traditional convolutional layer is replaced by using iconv in the backbone network; second, based on ShuffleNetV2, the MS attention mechanism module is used to improve the network structure, and a new backbone network structure isnet is proposed to reduce the computation and improve the accuracy of the detector while ensuring that the feature fusion performance is not affected. In the neck, we propose a GSPAN structure, firstly, we use 1*1 odconv to make the number of channels of features consistent with the minimum number of channels of backbone network output, which can effectively enhance the feature extraction ability of the network and reduce the network parameters; in addition, we down sample the GSPAN again and add a new feature scale to help the detector detect more targets and solve the problem of false detection due to occlusion and overlap caused by false detection, missed detection, and insufficient feature extraction ability. The experimental results show that the L-YOLOv5 model on RTX3090 is 44.3FPS, which is 82% less than the parameter amount of YOLOv5 and 3.3% more accurate on average, achieving excellent performance and satisfying the real-time detection of construction workers wearing helmets in construction scenarios of power systems.

Typical drug discovery and development processes are costly, time consuming and often biased by expert opinion. Aptamers are short, single-stranded oligonucleotides (RNA/DNA) that bind to target proteins and other types of biomolecules. Compared with small-molecule drugs, aptamers can bind to their targets with high affinity (binding strength) and specificity (uniquely interacting with the target only). The conventional development process for aptamers utilizes a manual process known as Systematic Evolution of Ligands by Exponential Enrichment (SELEX), which is costly, slow, dependent on library choice and often produces aptamers that are not optimized. To address these challenges, in this research, we create an intelligent approach, named DAPTEV, for generating and evolving aptamer sequences to support aptamer-based drug discovery and development. Using the COVID-19 spike protein as a target, our computational results suggest that DAPTEV is able to produce structurally complex aptamers with strong binding affinities.

As a methodology, Kansei Engineering (KE) has been developed to deal with consumers' subjective impressions and images of a product into the design elements of the product. One central step in KE is to generate Kansei profiles of the product. Traditional approaches to generating Kansei profiles, the average data model and voting based model, cannot model the underlying vagueness of Kansei data, in other words, they assume that any neighboring Kansei data have no semantic overlapping. This paper proposes a novel approach to generating Kansei profiles, which results with a probability distribution on Kansei data. The main advantage of our proposed approach is its ability to deal with partial semantic overlapping among Kansei data. The generated Kansei profiles can also be applied to consumer-oriented Kansei evaluation problems. This paper also discusses possible applications to fuzzy principal component analysis and fuzzy regression analysis.

The main focus of this paper is for decision analysis from the target-oriented point of view. Firstly, the target achievement computation method is revised, in which the resulting value function can have four shapes: concave, convex, S-shaped, inverse S-shaped. In addition, it is now more and more widely acknowledged that all facets of uncertainty cannot be captured by a single probability distribution. A fuzzy uncertain target-oriented method is also proposed, in which the proportional approach is selected to transform a possibility distribution into its associated probability distribution, and then based on the random target-oriented model, we can obtain the probability of meeting targets. Three types of fuzzy targets, widely used in Bellman-Zadeh paradigm, are selected to illustrate the fuzzy target-oriented model.