Genetic interactions reveal distinct biological and therapeutic implications in breast cancer.

Co-occurrence and mutual exclusivity of genomic alterations may reflect the existence of genetic interactions, potentially shaping distinct biological phenotypes and impacting therapeutic response in breast cancer. However, our understanding of them remains limited. Herein, we investigate a large-scale multi-omics cohort (n = 873) and a real-world clinical sequencing cohort (n = 4,405) including several clinical trials with detailed treatment outcomes and perform functional validation in patient-derived organoids, tumor fragments, and in vivo models. Through this comprehensive approach, we construct a network comprising co-alterations and mutually exclusive events and characterize their therapeutic potential and underlying biological basis. Notably, we identify associations between TP53mut-AURKAamp and endocrine therapy resistance, germline BRCA1mut-MYCamp and improved sensitivity to PARP inhibitors, and TP53mut-MYBamp and immunotherapy resistance. Furthermore, we reveal that precision treatment strategies informed by co-alterations hold promise to improve patient outcomes. Our study highlights the significance of genetic interactions in guiding genome-informed treatment decisions beyond single driver alterations.

The HLA-I landscape confers prognosis and antitumor immunity in breast cancer.

Breast cancer is a highly heterogeneous disease with varied subtypes, prognoses and therapeutic responsiveness. Human leukocyte antigen class I (HLA-I) shapes the immunity and thereby influences the outcome of breast cancer. However, the implications of HLA-I variations in breast cancer remain poorly understood. In this study, we established a multiomics cohort of 1156 Chinese breast cancer patients for HLA-I investigation. We calculated four important HLA-I indicators in each individual, including HLA-I expression level, somatic HLA-I loss of heterozygosity (LOH), HLA-I evolutionary divergence (HED) and peptide-binding promiscuity (Pr). Then, we evaluated their distribution and prognostic significance in breast cancer subtypes. We found that the four breast cancer subtypes had distinct features of HLA-I indicators. Increased expression of HLA-I and LOH were enriched in triple-negative breast cancer (TNBC), while Pr was relatively higher in hot tumors within TNBCs. In particular, a higher Pr indicated a better prognosis in TNBCs by regulating the infiltration of immune cells and the expression of immune molecules. Using the matched genomic and transcriptomic data, we found that mismatch repair deficiency-related mutational signature and pathways were enriched in low-Pr TNBCs, suggesting that targeting mismatch repair deficiency for synthetic lethality might be promising therapy for these patients. In conclusion, we presented an overview of HLA-I indicators in breast cancer and provided hints for precision treatment for low-Pr TNBCs.

Emergence of machine language: towards symbolic intelligence with neural networks.

Inspired by human language, machine language is a novel discrete representation learned from visual data only through playing the speak, guess, and draw game.

Integrated multiomic profiling of breast cancer in the Chinese population reveals patient stratification and therapeutic vulnerabilities.

Molecular profiling guides precision treatment of breast cancer; however, Asian patients are underrepresented in publicly available large-scale studies. We established a comprehensive multiomics cohort of 773 Chinese patients with breast cancer and systematically analyzed their genomic, transcriptomic, proteomic, metabolomic, radiomic and digital pathology characteristics. Here we show that compared to breast cancers in white individuals, Asian individuals had more targetable AKT1 mutations. Integrated analysis revealed a higher proportion of HER2-enriched subtype and correspondingly more frequent ERBB2 amplification and higher HER2 protein abundance in the Chinese HR+HER2+ cohort, stressing anti-HER2 therapy for these individuals. Furthermore, comprehensive metabolomic and proteomic analyses revealed ferroptosis as a potential therapeutic target for basal-like tumors. The integration of clinical, transcriptomic, metabolomic, radiomic and pathological features allowed for efficient stratification of patients into groups with varying recurrence risks. Our study provides a public resource and new insights into the biology and ancestry specificity of breast cancer in the Asian population, offering potential for further precision treatment approaches.

ZNF689 deficiency promotes intratumor heterogeneity and immunotherapy resistance in triple-negative breast cancer.

Triple-negative breast cancer (TNBC) is an aggressive disease characterized by remarkable intratumor heterogeneity (ITH), which poses therapeutic challenges. However, the clinical relevance and key determinant of ITH in TNBC are poorly understood. Here, we comprehensively characterized ITH levels using multi-omics data across our center's cohort (n = 260), The Cancer Genome Atlas cohort (n = 134), and four immunotherapy-treated cohorts (n = 109). Our results revealed that high ITH was associated with poor patient survival and immunotherapy resistance. Importantly, we identified zinc finger protein 689 (ZNF689) deficiency as a crucial determinant of ITH formation. Mechanistically, the ZNF689-TRIM28 complex was found to directly bind to the promoter of long interspersed element-1 (LINE-1), inducing H3K9me3-mediated transcriptional silencing. ZNF689 deficiency reactivated LINE-1 retrotransposition to exacerbate genomic instability, which fostered ITH. Single-cell RNA sequencing, spatially resolved transcriptomics and flow cytometry analysis confirmed that ZNF689 deficiency-induced ITH inhibited antigen presentation and T-cell activation, conferring immunotherapy resistance. Pharmacological inhibition of LINE-1 significantly reduced ITH, enhanced antitumor immunity, and eventually sensitized ZNF689-deficient tumors to immunotherapy in vivo. Consistently, ZNF689 expression positively correlated with favorable prognosis and immunotherapy response in clinical samples. Altogether, our study uncovers a previously unrecognized mechanism underlying ZNF689 deficiency-induced ITH and suggests LINE-1 inhibition combined with immunotherapy as a novel treatment strategy for TNBC.

Revisiting Confidence Estimation: Towards Reliable Failure Prediction.

Reliable confidence estimation is a challenging yet fundamental requirement in many risk-sensitive applications. However, modern deep neural networks are often overconfident for their incorrect predictions, i.e., misclassified samples from known classes, and out-of-distribution (OOD) samples from unknown classes. In recent years, many confidence calibration and OOD detection methods have been developed. In this paper, we find a general, widely existing but actually-neglected phenomenon that most confidence estimation methods are harmful for detecting misclassification errors. We investigate this problem and reveal that popular calibration and OOD detection methods often lead to worse confidence separation between correctly classified and misclassified examples, making it difficult to decide whether to trust a prediction or not. Finally, we propose to enlarge the confidence gap by finding flat minima, which yields state-of-the-art failure prediction performance under various settings including balanced, long-tailed, and covariate-shift classification scenarios. Our study not only provides a strong baseline for reliable confidence estimation but also acts as a bridge between understanding calibration, OOD detection, and failure prediction.

Distributed adaptive sliding-mode control for 2-D plane vehicle platoon with prescribed performance and angle constraint.

This paper focuses on the distributed adaptive sliding-mode control problem for two-dimensional (2-D) plane vehicle platoon with prescribed performance, angle constraints, and actuator faults. The quadratic spacing policy (QSP) is first adopted for the 2-D plane vehicle platoon to adjust the inter-vehicle spacing. The spacing error can converge within a finite time to the small region predetermined by a new finite-time performance function (FTPF). Meanwhile, a new transformed error function is introduced to convert the FTPF-constrained spacing errors into equivalent unconstrained ones. Besides, the property of the invertible nonlinear mapping function is used for the original system with the angle constraint to get a new unconstrained system. Moreover, a new controller based on hyperbolic tangent function is designed to handle actuator faults occurring multiple times over a period. Furthermore, the stability and string stability of the 2-D plane vehicle platoon are achieved through sliding-mode control. Finally, the simulation results validate the effectiveness of the proposed techniques.

Antigen-induced chimeric antigen receptor multimerization amplifies on-tumor cytotoxicity.

Ligand-induced receptor dimerization or oligomerization is a widespread mechanism for ensuring communication specificity, safeguarding receptor activation, and facilitating amplification of signal transduction across the cellular membrane. However, cell-surface antigen-induced multimerization (dubbed AIM herein) has not yet been consciously leveraged in chimeric antigen receptor (CAR) engineering for enriching T cell-based therapies. We co-developed ciltacabtagene autoleucel (cilta-cel), whose CAR incorporates two B-cell maturation antigen (BCMA)-targeted nanobodies in tandem, for treating multiple myeloma. Here we elucidated a structural and functional model in which BCMA-induced cilta-cel CAR multimerization amplifies myeloma-targeted T cell-mediated cytotoxicity. Crystallographic analysis of BCMA-nanobody complexes revealed atomic details of antigen-antibody hetero-multimerization whilst analytical ultracentrifugation and small-angle X-ray scattering characterized interdependent BCMA apposition and CAR juxtaposition in solution. BCMA-induced nanobody CAR multimerization enhanced cytotoxicity, alongside elevated immune synapse formation and cytotoxicity-mediating cytokine release, towards myeloma-derived cells. Our results provide a framework for contemplating the AIM approach in designing next-generation CARs.

Molecular classification of hormone receptor-positive HER2-negative breast cancer.

Hormone receptor-positive (HR+)/human epidermal growth factor receptor 2-negative (HER2-) breast cancer is the most prevalent type of breast cancer, in which endocrine therapy resistance and distant relapse remain unmet challenges. Accurate molecular classification is urgently required for guiding precision treatment. We established a large-scale multi-omics cohort of 579 patients with HR+/HER2- breast cancer and identified the following four molecular subtypes: canonical luminal, immunogenic, proliferative and receptor tyrosine kinase (RTK)-driven. Tumors of these four subtypes showed distinct biological and clinical features, suggesting subtype-specific therapeutic strategies. The RTK-driven subtype was characterized by the activation of the RTK pathways and associated with poor outcomes. The immunogenic subtype had enriched immune cells and could benefit from immune checkpoint therapy. In addition, we developed convolutional neural network models to discriminate these subtypes based on digital pathology for potential clinical translation. The molecular classification provides insights into molecular heterogeneity and highlights the potential for precision treatment of HR+/HER2- breast cancer.

Cycle-Consistent Weakly Supervised Visual Grounding With Individual and Contextual Representations.

Visual grounding, aiming to align image regions with textual queries, is a fundamental task for cross-modal learning. We study the weakly supervised visual grounding, where only image-text pairs at a coarse-grained level are available. Due to the lack of fine-grained correspondence information, existing approaches often encounter matching ambiguity. To overcome this challenge, we introduce the cycle consistency constraint into region-phrase pairs, which strengthens correlated pairs and weakens unrelated pairs. This cycle pairing makes use of the bidirectional association between image regions and text phrases to alleviate matching ambiguity. Furthermore, we propose a parallel grounding framework, where backbone networks and subsequent relation modules extract individual and contextual representations to calculate context-free and context-aware similarities between regions and phrases separately. Those two representations characterize visual/linguistic individual concepts and inter-relationships, respectively, and then complement each other to achieve cross-modal alignment. The whole framework is trained by minimizing an image-text contrastive loss and a cycle consistency loss. During inference, the above two similarities are fused to give the final region-phrase matching score. Experiments on five popular datasets about visual grounding demonstrate a noticeable improvement in our method. The source code is available at https://github.com/Evergrow/WSVG.

Genomic alterations affecting tumor-infiltrating lymphocytes and PD-L1 expression patterns in triple-negative breast cancer.

BACKGROUND: Tumor-infiltrating lymphocytes (TILs) and programmed cell death 1 ligand 1 (PD-L1) remain imperfect in predicting clinical outcomes of triple-negative breast cancer because outcomes do not always correlate with the expression of these biomarkers. Genomic and transcriptomic alterations that may contribute to the expression of these biomarkers remain incompletely uncovered. METHODS: We evaluated PD-L1 immunohistochemistry scores (SP142 and 28-8 assays) and TILs in our triple-negative breast cancer multiomics dataset and 2 immunotherapy clinical trial cohorts. Then, we analyzed genomic and transcriptomic alterations correlated with TILs, PD-L1 expression, and patient outcomes. RESULTS: Despite TILs serving as a decent predictor for triple-negative breast cancer clinical outcomes, exceptions remained. Our study revealed that several genomic alterations were correlated with unexpected events. In particular, PD-L1 expression may cause a paradoxical relationship between TILs and prognosis in certain patients. Consequently, we classified triple-negative breast cancers into 4 groups based on PD-L1 and TIL levels. The TIL-negative PD-L1-positive and TIL-positive PD-L1-negative groups were not typical "hot" tumors; both were associated with worse prognoses and lower immunotherapy efficacy than TIL-positive PD-L1-positive tumors. Copy number variation of PD-L1 and oncogenic signaling activation were correlated with PD-L1 expression in the TIL-negative PD-L1-positive group, whereas GSK3B-induced degradation may cause undetectable PD-L1 expression in the TIL-positive PD-L1-negative group. These factors have the potential to affect the predictive function of both PD-L1 and TILs. CONCLUSIONS: Several genomic and transcriptomic alterations may cause paradoxical effects among TILs, PD-L1 expression, and prognosis in triple-negative breast cancer. Investigating and targeting these factors will advance precision immunotherapy for patients with this disease.

A New Language-Independent Deep CNN for Scene Text Detection and Style Transfer in Social Media Images.

Due to the adverse effect of quality caused by different social media and arbitrary languages in natural scenes, detecting text from social media images and transferring its style is challenging. This paper presents a novel end-to-end model for text detection and text style transfer in social media images. The key notion of the proposed work is to find dominant information, such as fine details in the degraded images (social media images), and then restore the structure of character information. Therefore, we first introduce a novel idea of extracting gradients from the frequency domain of the input image to reduce the adverse effect of different social media, which outputs text candidate points. The text candidates are further connected into components and used for text detection via a UNet++ like network with an EfficientNet backbone (EffiUNet++). Then, to deal with the style transfer issue, we devise a generative model, which comprises a target encoder and style parameter networks (TESP-Net) to generate the target characters by leveraging the recognition results from the first stage. Specifically, a series of residual mapping and a position attention module are devised to improve the shape and structure of generated characters. The whole model is trained end-to-end so as to optimize the performance. Experiments on our social media dataset, benchmark datasets of natural scene text detection and text style transfer show that the proposed model outperforms the existing text detection and style transfer methods in multilingual and cross-language scenario.

Imitating the oracle: Towards calibrated model for class incremental learning.

Class-incremental learning (CIL) aims to recognize classes that emerged in different phases. The joint-training (JT), which trains the model jointly with all classes, is often considered as the upper bound of CIL. In this paper, we thoroughly analyze the difference between CIL and JT in feature space and weight space. Motivated by the comparative analysis, we propose two types of calibration: feature calibration and weight calibration to imitate the oracle (ItO), i.e., JT. Specifically, on the one hand, feature calibration introduces deviation compensation to maintain the class decision boundary of old classes in feature space. On the other hand, weight calibration leverages forgetting-aware weight perturbation to increase transferability and reduce forgetting in parameter space. With those two calibration strategies, the model is forced to imitate the properties of joint-training at each incremental learning stage, thus yielding better CIL performance. Our ItO is a plug-and-play method and can be implemented into existing methods easily. Extensive experiments on several benchmark datasets demonstrate that ItO can significantly and consistently improve the performance of existing state-of-the-art methods. Our code is publicly available at https://github.com/Impression2805/ItO4CIL.

Learning by Seeing More Classes.

Traditional pattern recognition models usually assume a fixed and identical number of classes during both training and inference stages. In this paper, we study an interesting but ignored question: can increasing the number of classes during training improve the generalization and reliability performance? For a k-class problem, instead of training with only these k classes, we propose to learn with k+m classes, where the additional m classes can be either real classes from other datasets or synthesized from known classes. Specifically, we propose two strategies for constructing new classes from known classes. By making the model see more classes during training, we can obtain several advantages. First, the added m classes serve as a regularization which is helpful to improve the generalization accuracy on the original k classes. Second, this will alleviate the overconfident phenomenon and produce more reliable confidence estimation for different tasks like misclassification detection, confidence calibration, and out-of-distribution detection. Lastly, the additional classes can also improve the learned feature representation, which is beneficial for new classes generalization in few-shot learning and class-incremental learning. Compared with the widely proved concept of data augmentation (dataAug), our method is driven from another dimension of augmentation based on additional classes (classAug). Comprehensive experiments demonstrated the superiority of our classAug under various open-environment metrics on benchmark datasets.

Mixed-Supervised Scene Text Detection With Expectation-Maximization Algorithm.

Scene text detection is an important and challenging task in computer vision. For detecting arbitrarily-shaped texts, most existing methods require heavy data labeling efforts to produce polygon-level text region labels for supervised training. In order to reduce the cost in data labeling, we study mixed-supervised arbitrarily-shaped text detection by combining various weak supervision forms (e.g., image-level tags, coarse, loose and tight bounding boxes), which are far easier to annotate. Whereas the existing weakly-supervised learning methods (such as multiple instance learning) do not promote full object coverage, to approximate the performance of fully-supervised detection, we propose an Expectation-Maximization (EM) based mixed-supervised learning framework to train scene text detector using only a small amount of polygon-level annotated data combined with a large amount of weakly annotated data. The polygon-level labels are treated as latent variables and recovered from the weak labels by the EM algorithm. A new contour-based scene text detector is also proposed to facilitate the use of weak labels in our mixed-supervised learning framework. Extensive experiments on six scene text benchmarks show that (1) using only 10% strongly annotated data and 90% weakly annotated data, our method yields comparable performance to that of fully supervised methods, (2) with 100% strongly annotated data, our method achieves state-of-the-art performance on five scene text benchmarks (CTW1500, Total-Text, ICDAR-ArT, MSRA-TD500, and C-SVT), and competitive results on the ICDAR2015 Dataset. We will make our weakly annotated datasets publicly available.

Unsupervised Structure-Texture Separation Network for Oracle Character Recognition.

Oracle bone script is the earliest-known Chinese writing system of the Shang dynasty and is precious to archeology and philology. However, real-world scanned oracle data are rare and few experts are available for annotation which make the automatic recognition of scanned oracle characters become a challenging task. Therefore, we aim to explore unsupervised domain adaptation to transfer knowledge from handprinted oracle data, which are easy to acquire, to scanned domain. We propose a structure-texture separation network (STSN), which is an end-to-end learning framework for joint disentanglement, transformation, adaptation and recognition. First, STSN disentangles features into structure (glyph) and texture (noise) components by generative models, and then aligns handprinted and scanned data in structure feature space such that the negative influence caused by serious noises can be avoided when adapting. Second, transformation is achieved via swapping the learned textures across domains and a classifier for final classification is trained to predict the labels of the transformed scanned characters. This not only guarantees the absolute separation, but also enhances the discriminative ability of the learned features. Extensive experiments on Oracle-241 dataset show that STSN outperforms other adaptation methods and successfully improves recognition performance on scanned data even when they are contaminated by long burial and careless excavation.

The early-stage triple-negative breast cancer landscape derives a novel prognostic signature and therapeutic target.

PURPOSE: Triple-negative breast cancer (TNBC) is a highly heterogeneous disease. Patients with early-stage TNBCs have distinct likelihood of distant recurrence. This study aimed to develop a prognostic signature of early-stage TNBC patients to improve risk stratification. METHODS: Using RNA-sequencing data, we analyzed 189 pathologically confirmed pT1-2N0M0 TNBC patients and identified 21 mRNAs that were highly expressed in tumor and related to relapse-free survival. All-subset regression program was used for constructing a 7-mRNA signature in the training set (n = 159); the accuracy and prognostic value were then validated using an independent validation set (n = 158). RESULTS: Here, we profiled the transcriptome data from 189 early-stage TNBC patients along with 50 paired normal tissues. Early-stage TNBCs mainly consisted of basal-like immune-suppressed subtype and had higher homologous recombination deficiency scores. We developed a prognostic signature including seven mRNAs (ACAN, KRT5, TMEM101, LCA5, RPP40, LAGE3, CDKL2). In both the training (n = 159) and validation set (n = 158), this signature could identify patients with relatively high recurrence risks and served as an independent prognostic factor. Time-dependent receiver operating curve showed that the signature had better prognostic value than traditional clinicopathological features in both sets. Functionally, we showed that TMEM101 promoted cell proliferation and migration in vitro, which represented a potential therapeutic target. CONCLUSIONS: Our 7-mRNA signature could accurately predict recurrence risks of early-stage TNBCs. This model may facilitate personalized therapy decision-making for early-stage TNBCs individuals.

Adaptive neural networks-based fixed-time fault-tolerant consensus tracking for uncertain multiple Euler-Lagrange systems.

This article addresses the fixed-time fault-tolerant consensus tracking (FTCT) problem for uncertain multiple Euler-Lagrange systems (MELS) with the digraph and actuator faults. Firstly, a fixed-time distributed observer (DO) is built to estimate the states of leader. Then, the approximation ability of radical basic function neural networks (RBFNN) is exploited to deal with the system uncertainties. By using backstepping technique, the novel fault-tolerant local control protocol (FTLCP) and updating laws are designed to ensure that error variables converge to the small adjacent area of zero within fixed-time. Eventually, the effectiveness and practicality of the presented method are demonstrated through a typical MELS simulation.

Deep Neural Network Self-Distillation Exploiting Data Representation Invariance.

To harvest small networks with high accuracies, most existing methods mainly utilize compression techniques such as low-rank decomposition and pruning to compress a trained large model into a small network or transfer knowledge from a powerful large model (teacher) to a small network (student). Despite their success in generating small models of high performance, the dependence of accompanying assistive models complicates the training process and increases memory and time cost. In this article, we propose an elegant self-distillation (SD) mechanism to obtain high-accuracy models directly without going through an assistive model. Inspired by the invariant recognition in the human vision system, different distorted instances of the same input should possess similar high-level data representations. Thus, we can learn data representation invariance between different distorted versions of the same sample. Especially, in our learning algorithm based on SD, the single network utilizes the maximum mean discrepancy metric to learn the global feature consistency and the Kullback-Leibler divergence to constrain the posterior class probability consistency across the different distorted branches. Extensive experiments on MNIST, CIFAR-10/100, and ImageNet data sets demonstrate that the proposed method can effectively reduce the generalization error for various network architectures, such as AlexNet, VGGNet, ResNet, Wide ResNet, and DenseNet, and outperform existing model distillation methods with little extra training efforts.

Meta-Prototypical Learning for Domain-Agnostic Few-Shot Recognition.

Few-shot learning (FSL) aims to classify novel images based on a few labeled samples with the help of meta-knowledge. Most previous works address this problem based on the hypothesis that the training set and testing set are from the same domain, which is not realistic for some real-world applications. Thus, we extend FSL to domain-agnostic few-shot recognition, where the domain of the testing task is unknown. In domain-agnostic few-shot recognition, the model is optimized on data from one domain and evaluated on tasks from different domains. Previous methods for FSL mostly focus on learning general features or adapting to few-shot tasks effectively. They suffer from inappropriate features or complex adaptation in domain-agnostic few-shot recognition. In this brief, we propose meta-prototypical learning to address this problem. In particular, a meta-encoder is optimized to learn the general features. Different from the traditional prototypical learning, the meta encoder can effectively adapt to few-shot tasks from different domains by the traces of the few labeled examples. Experiments on many datasets demonstrate that meta-prototypical learning performs competitively on traditional few-shot tasks, and on few-shot tasks from different domains, meta-prototypical learning outperforms related methods.