A novel homozygous Y140X mutation of ISG15 causes diverse type I interferonopathies in sibling patients with cutaneous lesions or recurrent parenchymal pneumonia.

PURPOSE: Interferon-stimulated gene 15 (ISG15) deficiency, a rare human inborn error of immunity characterized by susceptibility to Bacillus Calmette-Guerin (BCG) diseases, neuropathic and dermatological manifestations. METHODS: The clinical and immunological features of two siblings with ISG15 deficiency combined with asymptomatic myeloperoxidase (MPO) mutations were analyzed, and their pathogenesis, as well as target therapeutic candidates, were explored. RESULTS: The manifestation in patient 2 was skin lesions, while those in patient 1 were intracranial calcification and recurrent pneumonia. Whole-exome identified novel, dual mutations in ISG15 and MPO. PBMCs and B cell lines derived from the patients showed hyper-activated JAK/STAT signaling. Normal neutrophil function excluded pathogenicity caused by the MPO mutation. RNA sequencing identified baricitinib as therapeutic candidate. CONCLUSIONS: We report two sibling patients harboring the same novel ISG15 mutation showing diverse clinical features, and one harbored a rare phenotype of pneumonia. These findings expand the clinical spectrum of ISG15 deficiency and identify baricitinib as therapeutic candidate.

An early-onset SLE patient with a novel paternal inherited BACH2 mutation.

BACH2-related immunodeficiency and autoimmunity (BRIDA) is an inborn error of immunity, newly reported in 2017, presenting with symptoms of immunoglobulin deficiency and ongoing colitis. Studies using a mouse model have demonstrated that BACH2 deficiency predisposes individuals to systemic lupus erythematosus (SLE); however, no BACH2 deficiency has been reported in SLE patients. Here we describe a patient with BRIDA presenting with early-onset SLE, juvenile dermatomyositis, and IgA deficiency. Whole exome sequencing analysis of the patient and her parents revealed a novel heterozygous point mutation in BACH2, c.G1727T, resulting in substitution of a highly conserved arginine with leucine (R576L), which is predicted to be deleterious, in the patient and her father. Reduced BACH2 expression and deficient transcriptional repression of the BACH2 target, BLIMP1, were detected in PBMCs or lymphoblastoid cell lines of our patient. Notably, extreme reduction of memory B cells was detected in the patient's father, although he had no obvious symptoms. SLE symptoms and recurrent fever were relieved by treatment with prednisone combined with tofacitinib. Thus, we present the second report of BRIDA and demonstrate that BACH2 may be a monogenic cause of SLE.

InOR-Net: Incremental 3-D Object Recognition Network for Point Cloud Representation.

3-D object recognition has successfully become an appealing research topic in the real world. However, most existing recognition models unreasonably assume that the categories of 3-D objects cannot change over time in the real world. This unrealistic assumption may result in significant performance degradation for them to learn new classes of 3-D objects consecutively due to the catastrophic forgetting on old learned classes. Moreover, they cannot explore which 3-D geometric characteristics are essential to alleviate the catastrophic forgetting on old classes of 3-D objects. To tackle the above challenges, we develop a novel Incremental 3-D Object Recognition Network (i.e., InOR-Net), which could recognize new classes of 3-D objects continuously by overcoming the catastrophic forgetting on old classes. Specifically, category-guided geometric reasoning is proposed to reason local geometric structures with distinctive 3-D characteristics of each class by leveraging intrinsic category information. We then propose a novel critic-induced geometric attention mechanism to distinguish which 3-D geometric characteristics within each class are beneficial to overcome the catastrophic forgetting on old classes of 3-D objects while preventing the negative influence of useless 3-D characteristics. In addition, a dual adaptive fairness compensations' strategy is designed to overcome the forgetting brought by class imbalance by compensating biased weights and predictions of the classifier. Comparison experiments verify the state-of-the-art performance of the proposed InOR-Net model on several public point cloud datasets.

Open-Ended Online Learning for Autonomous Visual Perception.

The visual perception systems aim to autonomously collect consecutive visual data and perceive the relevant information online like human beings. In comparison with the classical static visual systems focusing on fixed tasks (e.g., face recognition for visual surveillance), the real-world visual systems (e.g., the robot visual system) often need to handle unpredicted tasks and dynamically changed environments, which need to imitate human-like intelligence with open-ended online learning ability. Therefore, we provide a comprehensive analysis of open-ended online learning problems for autonomous visual perception in this survey. Based on "what to online learn" among visual perception scenarios, we classify the open-ended online learning methods into five categories: instance incremental learning to handle data attributes changing, feature evolution learning for incremental and decremental features with the feature dimension changed dynamically, class incremental learning and task incremental learning aiming at online adding new coming classes/tasks, and parallel and distributed learning for large-scale data to reveal the computational and storage advantages. We discuss the characteristic of each method and introduce several representative works as well. Finally, we introduce some representative visual perception applications to show the enhanced performance when using various open-ended online learning models, followed by a discussion of several future directions.

Multiple Immune Defects in Two Patients with Novel DOCK2 Mutations Result in Recurrent Multiple Infection Including Live Attenuated Virus Vaccine.

The dedicator of cytokinesis 2(DOCK2) protein, an atypical guanine nucleotide exchange factor (GEFs), is a member of the DOCKA protein subfamily. DOCK2 protein deficiency is characterized by early-onset lymphopenia, recurrent infections, and lymphocyte dysfunction, which was classified as combined immune deficiency with neutrophil abnormalities as well. The only cure is hematopoietic stem cell transplantation. Here, we report two patients harboring four novel DOCK2 mutations associated with recurrent infections including live attenuated vaccine-related infections. The patient's condition was partially alleviated by symptomatic treatment or intravenous immunoglobulin. We also confirmed defects in thymic T cell output and T cell proliferation, as well as aberrant skewing of T/B cell subset TCR-Vß repertoires. In addition, we noted neutrophil defects, the weakening of actin polymerization, and BCR internalization under TCR/BCR activation. Finally, we found that the DOCK2 protein affected antibody affinity although with normal total serum immunoglobulin. The results reported herein expand the clinical phenotype, the pathogenic DOCK2 mutation database, and the immune characteristics of DOCK2-deficient patients.

A Multicenter Cohort Study of Immune Dysregulation Disorders Caused by ELF4 Variants in China.

Patients with DEX (deficiency in ELF4, X-linked) were recently reported by our team and others, and cases are very limited worldwide. Our knowledge of this new disease is currently preliminary. In this study, we described 5 more cases presenting mainly with oral ulcer, inflammatory bowel disease-like symptoms, fever of unknown origin, anemia, or systemic lupus erythematosus. Whole exome sequencing identified potential pathogenic ELF4 variants in all cases. The pathogenicity of these variants was confirmed by the detection of ELF4 expression in peripheral blood mononuclear cells from patients and utilizing a simple IFN-b luciferase reporter assay, as previously reported. Our findings significantly contribute to the current understanding of DEX.

De Novo Somatic Mosaicism of CYBB Caused by Intronic LINE-1 Element Insertion Resulting in Chronic Granulomatous Disease.

Chronic granulomatosis disease (CGD) is a rare inborn error of immunity, characterized by phagocytic respiratory outbreak dysfunction. Mutations causing CGD occur in CYBB on the X chromosome and in the autosomal genes CYBA, NCF1, NCF2, NCF4, RAC2, and CYBC1. Nevertheless, some patients are clinically diagnosed with CGD, due to abnormal respiratory outbursts, while the pathogenic gene mutation is unidentified. Here, we report a patient with CGD who first presented with Bacillus Calmette-Guérin disease and had recurrent pneumonia. He was diagnosed with CGD by nitro blue tetrazolium and respiratory burst tests. Detailed assessment of neutrophil activity revealed that patient neutrophils were almost entirely nonfunctional. Sanger sequencing detected a 6-kb insertion of a LINE-1 transposable element in the third intron of CYBB, leading to abnormal splicing and pseudoexon insertion, as well as introduction of a premature termination codon, resulting in predicted protein truncation. Clonal analysis demonstrated that the patient had somatic mosaicism, and the phagocytes were almost all variant CYBB, while the mosaicism rate of PBMC was about 65%. Finally, deep RNA sequencing and gp91phox expression analysis confirmed the pathogenicity of the mutation. In conclusion, we demonstrate that insertion of a LINE-1 transposon in a CYBB intron was responsible for CGD in our patient. Intron LINE-1 transposon element insertion should be examined in CGD patients without any known disease-causing gene mutation, in addition to identification of new genes.

PFDN: Pyramid Feature Decoupling Network for Single Image Deraining.

Restoring images degraded by rain has attracted more academic attention since rain streaks could reduce the visibility of outdoor scenes. However, most existing deraining methods attempt to remove rain while recovering details in a unified framework, which is an ideal and contradictory target in the image deraining task. Moreover, the relative independence of rain streak features and background features is usually ignored in the feature domain. To tackle these challenges above, we propose an effective Pyramid Feature Decoupling Network (i.e., PFDN) for single image deraining, which could accomplish image deraining and details recovery with the corresponding features. Specifically, the input rainy image features are extracted via a recurrent pyramid module, where the features for the rainy image are divided into two parts, i.e., rain-relevant and rain-irrelevant features. Afterwards, we introduce a novel rain streak removal network for rain-relevant features and remove the rain streak from the rainy image by estimating the rain streak information. Benefiting from lateral outputs, we propose an attention module to enhance the rain-irrelevant features, which could generate spatially accurate and contextually reliable details for image recovery. For better disentanglement, we also enforce multiple causality losses at the pyramid features to encourage the decoupling of rain-relevant and rain-irrelevant features from the high to shallow layers. Extensive experiments demonstrate that our module can well model the rain-relevant information over the domain of the feature. Our framework empowered by PFDN modules significantly outperforms the state-of-the-art methods on single image deraining with multiple widely-used benchmarks, and also shows superiority in the fully-supervised domain.

Bi-ventricular assessment with cardiovascular magnetic resonance at 5 Tesla: A pilot study.

Background: Cardiovascular magnetic resonance (CMR) imaging at ultra-high fields (UHF) such as 7T has encountered many challenges such as faster T 2 * relaxation, stronger B0 and B1+ field inhomogeneities and additional safety concerns due to increased specific absorption rate (SAR) and peripheral nervous stimulation (PNS). Recently, a new line of 5T whole body MRI system has become available, and this study aims at evaluating the performance and benefits of this new UHF system for CMR imaging. Methods: Gradient echo (GRE) CINE imaging was performed on healthy volunteers at both 5 and 3T, and was compared to balanced steady-state-free-procession (bSSFP) CINE imaging at 3T as reference. Higher spatial resolution GRE CINE scans were additionally performed at 5T. All scans at both fields were performed with ECG-gating and breath-holding. Image quality was blindly evaluated by two radiologists, and the cardiac functional parameters (e.g., EDV/ESV/mass/EF) of the left and right ventricles were measured for statistical analyses using the Wilcoxon signed-rank test and Bland-Altman analysis. Results: Compared to 3T GRE CINE imaging, 5T GRE CINE imaging achieved comparable or improved image quality with significantly superior SNR and CNR, and it has also demonstrated excellent capability for high resolution (1.0 × 1.0 × 6.0 mm3) imaging. Functional assessments from 5T GRE CINE images were highly similar with the 3T bSSFP CINE reference. Conclusions: This pilot study has presented the initial evaluation of CMR CINE imaging at 5T UHF, which yielded superior image quality and accurate functional quantification when compared to 3T counterparts. Along with reliable ECG gating, the new 5T UHF system has the potential to achieve well-balanced performance for CMR applications.

Clinical and immunological characteristics of five patients with immune dysregulation, polyendocrinopathy, enteropathy, X-linked syndrome in China-expanding the atypical phenotypes.

Background: Immune dysregulation, polyendocrinopathy, enteropathy, X-linked (IPEX) syndrome is a rare disorder of the immune regulatory system caused by forkhead box P3 (FOXP3) mutations. Abnormal numbers or functions of regulatory T (Treg) cells account for the various autoimmune symptoms. We aimed to explore the molecular genetics and phenotypic spectra of patients with atypical IPEX syndrome in China. Methods: We analyzed the molecular, clinical and immune phenotype characteristics of five Chinese patients with FOXP3 mutations. Results: We summarized the molecular and phenotypic features of five patients with FOXP3 mutations, including two novel mutations. Four of the five patients displayed atypical phenotypes, and one developed immune-related peripheral neuropathy. Three of the five patients showed normal frequencies of Treg cells, but the proportions of subsets of Treg cells, CD4+ T cells and B cells were out of balance. Conclusions: Our report broadens the understanding of the clinical features of atypical IPEX syndrome. Our detailed analyses of the immunological characteristics of these patients enhance the understanding of the possible mechanisms underlying the clinical manifestations.

Impact of Low-Head Dam Removal on River Morphology and Habitat Suitability in Mountainous Rivers.

Dam removal is considered an effective measure to solve the adverse ecological effects caused by dam construction and has started to be considered in China. The sediment migration and habitat restoration of river ecosystems after dam removal have been extensively studied abroad but are still in the exploratory stage in China. However, there are few studies on the ecological response of fishes at different growth stages. Considering the different habitat preferences of Schizothorax prenanti (S. prenanti) in the spawning and juvenile periods, this study coupled field survey data and a two-dimensional hydrodynamic model to explore the changes in river morphology at different scales and the impact of changes in hydrodynamic conditions on fish habitat suitability in the short term. The results show that after the dam is removed, in the upstream of the dam, the riverbed is eroded and cut down and the riverbed material coarsens. With the increase in flow velocity and the decrease in flow area, the weighted usable area (WUA) in the spawning and juvenile periods decreases by 5.52% and 16.36%, respectively. In the downstream of the dam, the riverbed is markedly silted and the bottom material becomes fine. With the increase in water depth and flow velocity, the WUA increases by 79.91% in the spawning period and decreases by 67.90% in the juvenile period, which is conducive to adult fish spawning but not to juvenile fish growth. The changes in physical habitat structure over a short time period caused by dam removal have different effects on different fish development periods, which are not all positive. The restoration of stream continuity increases adult fish spawning potential while limiting juvenile growth. Thus, although fish can spawn successfully, self-recruitment of fish stocks can still be affected if juvenile fish do not grow successfully. This study provides a research basis for habitat assessment after dam removal and a new perspective for the subsequent adaptive management strategy of the project.

Genetic and Functional Identifying of Novel STAT1 Loss-of-Function Mutations in Patients with Diverse Clinical Phenotypes.

PURPOSE: Mutations in signal transducer and activator of transcription 1 (STAT1) cause a broad spectrum of disease phenotypes. Heterozygous STAT1 loss-of-function (LOF) mutations cause Mendelian susceptibility to mycobacterial diseases (MSMD) infection, which is attributable to impaired IFN-γ signaling. The identification of novel mutations may extend the phenotypes associated with autosomal dominant (AD) STAT1 deficiency. METHODS: Five patients with heterozygous STAT1 variations were recruited and their clinical and immunologic phenotypes were analyzed, with particular reference to JAK-STAT1 signaling pathways. RESULTS: Four, heterozygous STAT1 deficiency mutations were identified, three of which were novel mutations. Two of the mutations were previously unreported mRNA splicing mutations in AD STAT1-deficient patients. Patients with heterozygous STAT1 deficiency suffered not only mycobacterial infection, but also intracellular non-mycobacterial bacterial infection and congenital multiple malformations. AD-LOF mutation impaired IFN-γ-mediated STAT1 phosphorylation, gamma-activated sequence (GAS), and IFN-stimulated response element (ISRE) transcription activity and IFN-induced gene expression to different extents, which might account for the diverse clinical manifestations observed in these patients. CONCLUSION: The infectious disease susceptibility and phenotypic spectrum of patients with AD STAT1-LOF are broader than simply MSMD. The susceptibility to infections and immunological deficiency phenotypes, observed in AD-LOF patients, confirms the importance of STAT1 in host-pathogen interaction and immunity. However, variability in the nature and extent of these phenotypes suggests that functional analysis is required to identify accurately novel, heterozygous STAT1 mutations, associated with pathogenicity. Aberrant splice of STAT1 RNA could result in AD-LOF for STAT1 signaling which need more cases for confirmation.

A contrastive learning approach for ICU false arrhythmia alarm reduction.

The high rate of false arrhythmia alarms in Intensive Care Units (ICUs) can lead to disruption of care, negatively impacting patients' health through noise disturbances, and slow staff response time due to alarm fatigue. Prior false-alarm reduction approaches are often rule-based and require hand-crafted features from physiological waveforms as inputs to machine learning classifiers. Despite considerable prior efforts to address the problem, false alarms are a continuing problem in the ICUs. In this work, we present a deep learning framework to automatically learn feature representations of physiological waveforms using convolutional neural networks (CNNs) to discriminate between true vs. false arrhythmia alarms. We use Contrastive Learning to simultaneously minimize a binary cross entropy classification loss and a proposed similarity loss from pair-wise comparisons of waveform segments over time as a discriminative constraint. Furthermore, we augment our deep models with learned embeddings from a rule-based method to leverage prior domain knowledge for each alarm type. We evaluate our method using the dataset from the 2015 PhysioNet Computing in Cardiology Challenge. Ablation analysis demonstrates that Contrastive Learning significantly improves the performance of a combined deep learning and rule-based-embedding approach. Our results indicate that the final proposed deep learning framework achieves superior performance in comparison to the winning entries of the Challenge.

Loss of Function Mutation in ELF4 Causes Autoinflammatory and Immunodeficiency Disease in Human.

Monogenic autoinflammatory diseases (mAIDs) are a heterogeneous group of diseases affecting primarily innate immunity, with various genetic causes. Genetic diagnosis of mAIDs can assist in the patient's management and therapy. However, a large number of sporadic and familial cases remain genetically uncharacterized. Deficiency in ELF4, X-linked (DEX) is recently identified as a novel mAID. Here, we described a pediatric patient suffering from recurrent viral and bacterial respiratory infection, refractory oral ulcer, constipation, and arthritis. Whole-exome sequencing found a hemizygous variant in ELF4 (chrX:129205133 A > G, c.691 T > C, p.W231R). Using cells from patient and point mutation mice, we showed mutant cells failed to restrict viral replication effectively and produced more pro-inflammatory cytokines. RNA-seq identified several potential critical antiviral and anti-inflammation genes with decreased expression, and ChIP-qPCR assay suggested mutant ELF4 failed to bind to the promoters of these genes. Thus, we presented the second report of DEX.

Representative Task Self-Selection for Flexible Clustered Lifelong Learning.

Consider the lifelong machine learning paradigm whose objective is to learn a sequence of tasks depending on previous experiences, e.g., knowledge library or deep network weights. However, the knowledge libraries or deep networks for most recent lifelong learning models are of prescribed size and can degenerate the performance for both learned tasks and coming ones when facing with a new task environment (cluster). To address this challenge, we propose a novel incremental clustered lifelong learning framework with two knowledge libraries: feature learning library and model knowledge library, called Flexible Clustered Lifelong Learning (FCL3). Specifically, the feature learning library modeled by an autoencoder architecture maintains a set of representation common across all the observed tasks, and the model knowledge library can be self-selected by identifying and adding new representative models (clusters). When a new task arrives, our FCL3 model firstly transfers knowledge from these libraries to encode the new task, i.e., effectively and selectively soft-assigning this new task to multiple representative models over feature learning library. Then: 1) the new task with a higher outlier probability will be judged as a new representative, and used to redefine both feature learning library and representative models over time; or 2) the new task with lower outlier probability will only refine the feature learning library. For model optimization, we cast this lifelong learning problem as an alternating direction minimization problem as a new task comes. Finally, we evaluate the proposed framework by analyzing several multitask data sets, and the experimental results demonstrate that our FCL3 model can achieve better performance than most lifelong learning frameworks, even batch clustered multitask learning models.

Novel mutations of TYK2 leading to divergent clinical phenotypes.

BACKGROUND: TYK2 deficiency is a rare primary immunodeficiency disease caused by loss-of-function mutations of TYK2 gene, which is initially proposed as a subset of hyper-IgE syndrome (HIES). However, accumulating evidence suggests TYK2-deficient patients do not necessarily present with HIES characteristics, indicating a vacuum of knowledge on the exact roles of TYK2 in human immune system. METHOD: Pathogenic effects of patients were confirmed by qRT-PCR, Western blot, and protein stability assays. The responses to cytokines including IFN-α/ß/γ, IL-6, IL-10, IL-12, and IL-23 of peripheral blood mononuclear cells (PBMCs) from these patients were detected by Western blot, qRT-PCR, and flow cytometry. The differentiation of T and B cells was detected by flow cytometry. RESULTS: We described five more TYK2-deficient cases presenting with or without hyper-IgE levels, atopy, and distinct pathogen infection profile, which are caused by novel TYK2 mutations. These mutations were all found by high-throughput sequencing and confirmed by Sanger sequencing. The patients showed heterogeneous responses to various cytokine treatments, including IFN-α/ß/γ, IL-6, IL-10, IL-12, and IL-23. The homeostasis of lymphocytes is also disrupted. CONCLUSION: Based on our findings, we propose that TYK2 works as a multi-tasker in orchestrating various cytokine signaling pathways, differentially combined defects which account for the expressed clinical manifestations.

Visual-Tactile Fused Graph Learning for Object Clustering.

Object clustering has received considerable research attention most recently. However, 1) most existing object clustering methods utilize visual information while ignoring important tactile modality, which would inevitably lead to model performance degradation and 2) simply concatenating visual and tactile information via multiview clustering method can make complementary information to not be fully explored, since there are many differences between vision and touch. To address these issues, we put forward a graph-based visual-tactile fused object clustering framework with two modules: 1) a modality-specific representation learning module MR and 2) a unified affinity graph learning module MU . Specifically, MR focuses on learning modality-specific representations for visual-tactile data, where deep non-negative matrix factorization (NMF) is adopted to extract the hidden information behind each modality. Meanwhile, we employ an autoencoder-like structure to enhance the robustness of the learned representations, and two graphs to improve its compactness. Furthermore, MU highlights how to mitigate the differences between vision and touch, and further maximize the mutual information, which adopts a minimizing disagreement scheme to guide the modality-specific representations toward a unified affinity graph. To achieve ideal clustering performance, a Laplacian rank constraint is imposed to regularize the learned graph with ideal connected components, where noises that caused wrong connections are removed and clustering labels can be obtained directly. Finally, we propose an efficient alternating iterative minimization updating strategy, followed by a theoretical proof to prove framework convergence. Comprehensive experiments on five public datasets demonstrate the superiority of the proposed framework.

Semi-Supervised Dual Relation Learning for Multi-Label Classification.

In a real-world scenario, an object could contain multiple tags instead of a single categorical label. To this end, multi-label learning (MLL) emerged. In MLL, the feature distributions are long-tailed and the complex semantic label relation and the long-tailed training samples are the main challenges. Semi-supervised learning is a potential solution. While, existing methods are mainly designed for single class scenario while ignoring the latent label relations. In addition, they cannot well handle the distribution shift commonly existing across source and target domains. To this end, a Semi-supervised Dual Relation Learning (SDRL) framework for multi-label classification is proposed. SDRL utilizes a few labeled samples as well as large scale unlabeled samples in the training stage. It jointly explores the inter-instance feature-level relation and the intra-instance label-level relation even from the unlabeled samples. In our model, a dual-classifier structure is deployed to obtain domain invariant representations. The prediction results from the classifiers are further compared and the most confident predictions are extracted as pseudo labels. A trainable label relation tensor is designed to explicitly explore the pairwise latent label relations and refine the predicted labels. SDRL is able to effectively and efficiently explore the feature-label relation as well as the label-label relation knowledge without any extra semantic knowledge. We evaluated SDRL in general and zero-shot multi-label classification tasks and we concluded that SDRL is superior to other SOTA baselines. Furthermore, extensive ablation studies have been done which reveal the effectiveness of each component in our framework.

Where and How to Transfer: Knowledge Aggregation-Induced Transferability Perception for Unsupervised Domain Adaptation.

Unsupervised domain adaptation without accessing expensive annotation processes of target data has achieved remarkable successes in semantic segmentation. However, most existing state-of-the-art methods cannot explore whether semantic representations across domains are transferable or not, which may result in the negative transfer brought by irrelevant knowledge. To tackle this challenge, in this paper, we develop a novel Knowledge Aggregation-induced Transferability Perception (KATP) for unsupervised domain adaptation, which is a pioneering attempt to distinguish transferable or untransferable knowledge across domains. Specifically, the KATP module is designed to quantify which semantic knowledge across domains is transferable, by incorporating transferability information propagation from global category-wise prototypes. Based on KATP, we design a novel KATP Adaptation Network (KATPAN) to determine where and how to transfer. The KATPAN contains a transferable appearance translation module T_A() and a transferable representation augmentation module T_R(), where both modules construct a virtuous circle of performance promotion. T_A() develops a transferability-aware information bottleneck to highlight where to adapt transferable visual characterizations and modality information; T_R() explores how to augment transferable representations while abandoning untransferable information, and promotes the translation performance of T_A() in return. Experiments on several representative datasets and a medical dataset support the state-of-the-art performance of our model.