Search

This paper investigates the issue of the event-triggered asynchronous dissipative control for Markov jump systems with unknown probabilities and packet losses. A hidden Markov model is used to describe the asynchronous problem. A Bernoulli model and an event-triggered scheme are adopted to address the packet loss and reduce the transmission rate of sampled signals, respectively. Furthermore, there exists unknown probability information of the controller. Then, Lyapunov functions are constructed to obtain sufficient conditions for the stochastic stability and dissipative performance of the closed-loop system. Moreover, a design method for controller parameters and event-triggered matrices is proposed. Finally, the effectiveness of the proposed approach is verified through a three-modal numerical example.

Abstract Background Medical Visual Question Answering (MedVQA) enhances diagnostic accuracy and healthcare delivery by leveraging artificial intelligence to interpret medical images. This study aims to redefine MedVQA as a generation task that mirrors human–machine interaction and to develop a model capable of integrating complex visual and textual information. Methods We constructed a large-scale medical visual-question answering dataset, PMC-VQA, containing 227,000 VQA pairs across 149,000 images that span various modalities and diseases. We introduced a generative model that aligns visual information from a pre-trained vision encoder with a large language model. This model was initially trained on PMC-VQA and subsequently fine-tuned on multiple public benchmarks. Results Here, we show that our model significantly outperforms existing MedVQA models in generating relevant, accurate free-form answers. We also propose a manually verified test set that presents a greater challenge and serves as a robust measure to monitor the advancement of generative MedVQA methods. Conclusions The PMC-VQA dataset proves to be an essential resource for the research community, and our model marks a significant breakthrough in MedVQA. We maintain a leaderboard to facilitate comprehensive evaluation and comparison, providing a centralized resource for benchmarking state-of-the-art approaches.

Abstract Background The combination of anti‐PD‐1 antibody serplulimab and chemotherapy is considered standard first‐line therapy for advanced esophageal squamous cell carcinoma (ESCC), but few later‐line treatments are available. Here we evaluated the therapeutic efficacy of the recombinant, humanized anti‐EGFR antibody HLX07 when used alone or together with serplulimab and chemotherapy against advanced ESCC. Methods This open‐label, non‐randomized, two‐cohort, phase 2 trial involved patients 18‐75 years old with histologically or cytologically confirmed locally advanced, unresectable, or metastatic ESCC, and an Eastern Cooperative Oncology Group performance status of 0‐1. Patients who had failed first‐line immuno‐chemotherapy or at least two lines of other systemic therapy received HLX07 monotherapy intravenously at a dose of 1,000 mg once every 2 weeks (Q2W). Patients with no prior systemic therapy received HLX07 (1,000 mg, day 1) and serplulimab (200 mg, day 1) intravenously Q2W for up to 2 years, concurrently with cisplatin (50 mg/m2, day 1) for up to 8 cycles and 5‐fluorouracil (1,200 mg/m2, days 1‐2) for up to 12 cycles intravenously Q2W. The primary endpoints were progression‐free survival (PFS) and objective response rate (ORR). Results Overall, 50 patients were enrolled. In the HLX07 monotherapy group, ORR was 15.0% (3/20), and the median PFS was 1.5 months (95% confidence interval [CI], 1.3 to 3.7). The median duration of response was not reached, and the rate of patients showing an objective response lasting at least 6 months was 66.7% (95% CI, 5.4 to 94.5). Two (10.0%, 2/20) patients experienced grade 3‐4 treatment‐related adverse events (TRAEs), including hypomagnesemia, hypocalcemia, and fatigue. No patient experienced grade 5 TRAEs. In the HLX07 combination group, the ORR was 60.0% (18/30), and the median PFS was 7.8 months (95% CI, 3.3 to 9.1). Fourteen (46.7%, 14/30) patients experienced grade 3‐4 TRAEs, and one (3.3%, 1/30) patient died due to serplulimab‐related pneumonitis. Conclusions HLX07 monotherapy and its combination with serplulimab and chemotherapy showed manageable toxicity and promising antitumor activity in patients with recurrent or metastatic ESCC. Randomized controlled trials are warranted to further establish the safety and efficacy of HLX07 against ESCC. Trial registration This trial was registered at Clinicaltrials.gov (NCT05221658).

Abstract Sacubitril/valsartan, a first-in-class angiotensin receptor neprilysin inhibitor, is widely used to treat heart failure. Despite its efficacy, sacubitril/valsartan inevitably causes adverse events such as hypotension, renal dysfunction, hyperkalemia, and angioedema. Sacubitril/valsartan-associated ototoxicity is often underreported in clinical studies and real-world settings. This study aimed to systematically characterize ototoxicity profiles using post-marketing data. Disproportionality analysis was performed using the Food and Drug Administration Adverse Event Reporting System (FAERS) database from 2015 to 2023. The reporting odds ratio (ROR) and information components (IC) were used to detect potential ototoxicity associated with sacubitril/valsartan. Significant signals were defined as ROR025 > 1 and IC025 > 0 in at least three reports. From 272,790 reports on sacubitril/valsartan and 10,798 (4.0%) reports on ototoxicity, sacubitril/valsartan and ototoxicity showed significant association (IC025 = 1.57; ROR025 = 2.99). Dizziness, hypoacusis, balance disorder, and deafness were detected as positive signals; hypoacusis was the strongest (IC025 = 3.21; ROR025 = 9.29). Median time to onset was 45 (10–99) days, and 62.4% of ototoxicity events occurred within 90 days of sacubitril/valsartan administration. Hospitalization accounted for 32.5% of ototoxic outcomes, while life-threatening events, death, and disabling accounted for only 3.9%, 3.4%, and 2.0%, respectively. Sacubitril/valsartan treatment increased ototoxicity risks. Therefore, promptly identifying and treating sacubitril/valsartan-related ototoxicity is critical.

The fretting wear of steel wire will reduce the service life of wire rope and severely threaten the safety of mine hoisting. To investigate the influence of torsion on the fretting wear behavior of steel wire, the fretting wear tests of steel wires with spiral contact structure under the action of tensile-torsional coupling forces were conducted, and the fatigue fracture failure mechanism was revealed. The result indicates that the CoF increases with increasing torsional angle. As the cycles increase, the sliding distance between the wires decreases, while the adhesion increases. The fretting area gradually transitions from gross sliding status to partial sliding status. The sliding distance and energy loss between the wires increase with increasing torsional angle. The wear depth and coefficient increase with increasing torsional angle, and the wear level of peak contact structure is more serious than that of valley contact structure. The primary wear mechanisms of steel wire are abrasive wear, adhesive wear and fatigue wear. The fatigue fractograph of steel wire is obviously divided into fatigue source zone, crack growth zone and final fracture zone. Abundant secondary cracks and dimples exist in the final fracture zone, and the fatigue fracture failure mechanism is mainly ductile fracture.

Abstract Alzheimer’s disease (AD) is a prevalent neurodegenerative disorder characterized by the accumulation of amyloid β protein (Aβ) and the hyper-phosphorylation of the microtubule-associated protein Tau. The ubiquitin-proteasome system (UPS) plays a pivotal role in determining the fate of proteins, and its dysregulation can contribute to the buildup of Aβ and Tau. Deubiquitinating enzymes (DUBs), working in conjunction with activating enzymes (E1), ubiquitin-conjugating enzymes (E2), and ubiquitin ligases (E3), actively maintain the delicate balance of protein homeostasis. DUBs specifically remove ubiquitin tags from proteins marked for degradation, thereby averting their proteasomal breakdown. Several DUBs have demonstrated their capacity to regulate the levels of Aβ and Tau by modulating their degree of ubiquitination, underscoring their potential as therapeutic targets for AD. In this context, we present a comprehensive review of AD-associated DUBs and elucidate their physiological roles. Moreover, we delve into the current advancements in developing inhibitors targeting these DUBs, including the determination of cocrystal structures with their respective targets. Additionally, we assess the therapeutic efficacy of these inhibitors in AD, aiming to establish a theoretical foundation for future AD treatments.

Abstract Developing a generalist radiology diagnosis system can greatly enhance clinical diagnostics. In this paper, we introduce RadDiag, a foundational model supporting 2D and 3D inputs across various modalities and anatomies, using a transformer-based fusion module for comprehensive disease diagnosis. Due to patient privacy concerns and the lack of large-scale radiology diagnosis datasets, we utilize high-quality, clinician-reviewed radiological images available online with diagnosis labels. Our dataset, RP3D-DiagDS, contains 40,936 cases with 195,010 scans covering 5568 disorders (930 unique ICD-10-CM codes). Experimentally, our RadDiag achieves 95.14% AUC on internal evaluation with the knowledge-enhancement strategy. Additionally, RadDiag can be zero-shot applied or fine-tuned to external diagnosis datasets sourced from various medical centers, demonstrating state-of-the-art results. In conclusion, we show that publicly shared medical data on the Internet is a tremendous and valuable resource that can potentially support building strong models for image understanding in healthcare.

Banana is a significant crop, and three banana leaf diseases, including Sigatoka, Cordana and Pestalotiopsis, have the potential to have a serious impact on banana production. Existing studies are insufficient to provide a reliable method for accurately identifying banana leaf diseases. Therefore, this paper proposes a novel method to identify banana leaf diseases. First, a new algorithm called K-scale VisuShrink algorithm (KVA) is proposed to denoise banana leaf images. The proposed algorithm introduces a new decomposition scale K based on the semi-soft and middle course thresholds, the ideal threshold solution is obtained and substituted with the newly established threshold function to obtain a less noisy banana leaf image. Then, this paper proposes a novel network for image identification called Ghost ResNeSt-Attention RReLU-Swish Net (GR-ARNet) based on Resnet50. In this, the Ghost Module is implemented to improve the network's effectiveness in extracting deep feature information on banana leaf diseases and the identification speed; the ResNeSt Module adjusts the weight of each channel, increasing the ability of banana disease feature extraction and effectively reducing the error rate of similar disease identification; the model's computational speed is increased using the hybrid activation function of RReLU and Swish. Our model achieves an average accuracy of 96.98% and a precision of 89.31% applied to 13,021 images, demonstrating that the proposed method can effectively identify banana leaf diseases.

Abstract Lipid droplets (LDs) feature a unique monolayer lipid membrane that has not been extensively studied due to the lack of suitable molecular probes that are able to distinguish this membrane from the LD lipid core. In this work, we present a three-pronged molecular probe design strategy that combines lipophilicity-based organelle targeting with microenvironment-dependent activation and design an LD membrane labeling pro-probe called LDM. Upon activation by the HClO/ClO− microenvironment that surrounds LDs, LDM pro-probe releases LDM-OH probe that binds to LD membrane proteins thus enabling visualization of the ring-like LD membrane. By utilizing LDM, we identify the dynamic mechanism of LD membrane contacts and their protein accumulation parameters. Taken together, LDM represents the first molecular probe for imaging LD membranes in live cells to the best of our knowledge, and represents an attractive tool for further investigations into the specific regulatory mechanisms with LD-related metabolism diseases and drug screening.

Abstract Antibody–drug conjugates (ADCs) are among the fastest-growing classes of anticancer drugs, making it crucial to evaluate their potential for causing peripheral neuropathy. We analyzed data from the FAERS database (January 1, 2014, to June 30, 2023) using disproportionality and Bayesian methods. We identified 3076 cases of ADC-associated peripheral neuropathy. Our study revealed significant signals for all ADCs (ROR 1.82, 95% CI 1.76–1.89). ADCs with tubulin-binding payloads showed significant peripheral neuropathy signals (ROR 2.31, 95% CI 2.23–2.40), whereas those with DNA-targeting (ROR 0.48, 95% CI 0.39–0.59) and topoisomerase 1 inhibitor (ROR 0.56, 95% CI 0.48–0.66) payloads exhibited non-significant signals. Signals for peripheral sensory neuropathy were 4.83, 2.44, 2.74, and 2.21 (calculated based on IC025) for brentuximab vedotin, trastuzumab emtansine, enfortumab vedotin, and polatuzumab vedotin, while signals for peripheral motor neuropathy were 5.31, 0.34, 2.27, and 0.03, respectively. The median time to onset for all ADCs was 127 days (interquartile range 40–457). Tisotumab vedotin had the highest hospitalization rate at 26.67%, followed by brentuximab vedotin at 25.5%. Trastuzumab emtansine had the highest mortality rate ,with 80 deaths (11.96%) among 669 cases. Based on FAERS database, only ADCs with tubulin-binding payloads exhibited significant peripheral neuropathy signals. Brentuximab vedotin and enfortumab vedotin showed similar profiles for peripheral sensory neuropathy and motor neuropathy. Given the delayed time to onset and potentially poor outcomes, ADC-related peripheral neuropathy warrants significant attention.

Cross-view image clustering (CIC) showcases immense potential in recognizing image patterns due to the power to aggregate information between views without labels. However, most CIC ignore intricate coupling relationships between category and redundant features in aggregating complementary information of cross-view images, which may restrict performance in recognizing patterns of images. To this end, a cross-view mutual information decoupling based deep generative clustering approach is proposed for recognizing image patterns (DMID-UIC), which contains information maximization deep generative module and self-supervised posterior inference module. Specifically, the former maximizes the mutual information between data and semantics within the generative adversarial network to decouple category and redundant features hidden in cross-view images, which guarantees the separation of different semantic distributions in the data space. The latter models the distribution fitting between generated data and the prior semantic code as a classification task, via treating partitioning results of common representations between views as self-supervised labels. Meanwhile, to better optimize the model, an EM optimization strategy is designed to enhance the above two module learning in an iterative manner. Finally, comprehensive results verify the superiority and effectiveness of DMID-UIC. DMID-UIC improves ACC by 10.1% on the Caltech 101-7 dataset, compared to the second-best result.

Abstract Background Combining immune checkpoint and proton pump inhibitors is widely used in cancer treatment. However, the drug-drug interactions of these substances are currently unknown. This study aimed to explore drug-drug interactions associated with concomitant immune checkpoint and proton pump inhibitors. Methods Data were obtained from the US Food and Drug Administration Adverse Event Reporting System from 2014 to 2023. Disproportionality analysis was used for data mining by calculating the reporting odds ratios (RORs) with 95% confidence intervals (95%Cls). The adjusted RORs (RORadj) were then analysed using logistic regression analysis, considering age, sex, and reporting year. Drug-drug interactions occur when a combination treatment enhances the frequency of an event. Further confirmation of the robustness of the findings was achieved using additive and multiplicative models, which are the two statistical methodologies for signal detection of DDIs using spontaneous reporting system. Results The total number of reports on immune checkpoint combined with proton pump inhibitors was 4,276. Median patient age was 66 years (interquartile range [IQR]: 60–74 years). Significant interaction signals were observed for congenital, familial and genetic disorders (RORadj = 2.66, 95%CI, 1.38–5.14, additive models = 0.7322, multiplicative models = 3.5142), hepatobiliary disorders (RORcrude = 6.64, 95%CI, 5.82–7.58, RORadj = 7.10, 95%CI, 6.16–8.18, additive models = 2.0525, multiplicative models = 1.1622), metabolism and nutrition disorders (RORcrude = 3.27, 95%CI, 2.90–3.69, RORadj = 2.66, 95%CI, 2.30–3.08, additive models = 0.6194), and skin and subcutaneous tissue disorders (RORcrude = 1.41, 95%CI, 1.26–1.58, RORadj = 1.53, 95%CI, 1.34–1.75, additive models = 0.6927, multiplicative models = 5.3599). Subset data analysis showed that programmed death-1 combined with proton pump inhibitors was associated with congenital, familial, and genetic disorders; hepatobiliary disorders; and skin and subcutaneous tissue disorders. Programmed death ligand-1 combined with proton pump inhibitors was associated with adverse reactions of metabolism and nutrition disorders. Cytotoxic T-lymphocyte antigen-4 combined with proton pump inhibitors was associated with congenital, familial, and genetic disorders, and skin and subcutaneous tissue disorders. Conclusions Based on real-world data, four Standardized MedDRA Query System Organ Class toxicities were identified as drug-drug interactions associated with combining immune checkpoint and proton pump inhibitors. Clinicians should be cautious when administering these drugs concomitantly. Preclinical trials and robust clinical studies are required to explore the mechanisms and relationships underlying interactions, thus improving understanding of drug-drug interactions associated with this combination therapy.

Abstract The development of open-source, multilingual medical language models can benefit a wide, linguistically diverse audience from different regions. To promote this domain, we present contributions from the following: First, we construct a multilingual medical corpus, containing approximately 25.5B tokens encompassing 6 main languages, termed as MMedC, enabling auto-regressive domain adaptation for general LLMs; Second, to monitor the development of multilingual medical LLMs, we propose a multilingual medical multi-choice question-answering benchmark with rationale, termed as MMedBench; Third, we have assessed a number of open-source large language models (LLMs) on our benchmark, along with those further auto-regressive trained on MMedC. Our final model, MMed-Llama 3, with only 8B parameters, achieves superior performance compared to all other open-source models on both MMedBench and English benchmarks, even rivaling GPT-4. In conclusion, in this work, We present a large-scale corpus, a benchmark and a series of models to support the development of multilingual medical LLMs.

Background: Antibody–drug conjugates (ADCs) combine the targeted nature of monoclonal antibodies with the potent efficacy of small-molecule cytotoxic drugs. However, they also carry unique safety risks, including lung toxicity. Objective: To conduct a systematic review and analysis of ADC-related interstitial lung disease (ILD) incidence, characteristics, and risk factors to optimize safe and effective clinical use. Design: ADC-related ILD reports from the FDA Adverse Event Reporting System (FAERS) database between January 2014 and March 2023 were analyzed. Methods: ADC-related ILD reports were retrieved from the FAERS database. Statistical analyses were conducted using reporting odds ratio (ROR) and information components (ICs). The lower limit of the 95% confidence interval (CI) was set for ROR (ROR025) >1 or IC (IC025) >0, and statistical significance was determined based on a minimum of three reports. Results: The study analyzed the statistical data on ADC-induced ILDs (1277 cases). Trastuzumab deruxtecan was reported to be the most frequent (38.4%). Among the 33 preferred terms (PTs) in standardized MedDRA queries (SMQ) = “Interstitial lung disease,” the three most common were as follows: ILD (40.6%), pneumonitis (27.9%), and acute respiratory distress syndrome (ARDS) (7.6%). Trastuzumab deruxtecan showed the strongest association with ILD (PT) and pneumonitis, whereas ARDS was associated with four different drugs. The median time to onset of ADC-related ILDs was 51 days (interquartile range (IQR), 16–196), with ARDS having the earliest median time to onset at 15 days (IQR, 6–52). The onsets of pneumonitis, ILD, lung infiltration, and pulmonary toxicity were similar. More than 26% of ADC-related ILD cases result in death, with ARDS having the highest mortality rate of 65.0%. Conclusion: ADCs are associated with an increased risk of pulmonary adverse events, such as ILDs, with significant differences between drugs and varying mortality rates for different adverse events, necessitating distinct monitoring and appropriate management.

In recent years, numerous deep learning-based methods have gained increasing attention in hyperspectral classification, particularly the Graph Neural Network, which exhibits superior capabilities in structural description. However, a single graph structure is not suitable for hyperspectral feature representation. Therefore, we propose a novel Multiple-Scale graph network structure, known as the Multi-Scale Dense Graph Attention network for hyperspectral classification. Firstly, semi-supervised local Fisher discriminant analysis and superpixel segmentation were employed for dimensionality reduction and multi-scale graph construction, respectively. Secondly, Spectral-Spatial convolution is applied to extract shallow features from the image. Subsequently, an improved graph self-attention network is sequentially applied to each scale graph, and the different scale graphs are densely connected through spatial feature alignment modules, designed using twice matrix multiplication. Finally, the combined pixel-level feature map from multiple graph spaces is derived, and Spectral-Spatial convolution is employed to fuse the abundant feature maps for hyperspectral classification. Experimental results on various hyperspectral datasets demonstrate the superiority of our MSDesGATnet over many state-of-the-art methods. The code is available at https://github.com/l7170/MSDesGAT.git.

Abstract In esophageal squamous cell carcinoma (ESCC), lymph node (LN) metastasis is associated with poor survival. Emerging evidence has demonstrated elevated CD8+ T‐cell levels in metastatic LNs following immunotherapy and increased chemoresistance. However, the underlying regulatory mechanisms of CD8+ T cells in chemoresistant/metastatic patients have not been elucidated. Given that metastasis is linked to aberrant splicing patterns, transcripts with alternative splicing forms also play a critical role. With spatial transcriptomics (ST), spatial isoform transcriptomics (SiT), single‐cell RNA sequencing (scRNA‐seq), and T‐cell receptor (TCR) sequencing, the spatial isoforms are analyzed quantitatively in human solid tumors and LNs. These isoforms are classified according to expression and spatial distribution patterns and proposed that the temporal heterogeneity in isoforms is attributed to isoform biogenesis dynamics. C1QC+ tumor‐associated macrophages (TAMs) contribute to the formation of metastases in the context of both immunotherapy and chemotherapy. Additionally, CD74 isoforms can be targeted by mRNA drugs, such as antisense oligonucleotides (ASOs) and RNA interference (RNAi), to prevent chemoresistance and metastasis. Overall, this work suggests that C1QC+ TAMs interact with CD8+ CXCL13+ Tex cells via enrichment with the CD74 isoform in the ESCC ‘s metastatic lymph node.

Abstract DNA nanotechnology plays a crucial role in precise cancer medicine. Currently, molecular logic circuits are applied to detect tumor‐specific biomarkers and control the release of therapeutic drugs. However, these systems lack self‐learning capabilities for intelligent diagnostics in biological samples, and their data processing capabilities are limited. Here, a molecular learning vector quantization neural network (LVQNN) model based on DNA strand displacement (DSD) technology for breast tumor diagnosis is developed. Compared to previous work, the molecular LVQNN boasts powerful computing abilities, handling high‐dimensional data for intelligent cancer diagnosis. To verify the feasibility and versatility of the network, two distinct typical datasets are selected: one from a single source with cell morphology data from 569 cases, and a more extensive one spanning different populations and ages, with miRNA gene expression data from 1881 cases. By using the molecular LVQNN, diagnostic experiments are conducted on 50 and 120 public individuals from these two datasets, respectively, achieving accuracy rates of 94% and 97.5%. This study demonstrates that the LVQNN model exhibits significant advantages in breast cancer diagnosis and enhances diagnostic accuracy while introducing new approaches for intelligent cancer diagnosis, anticipated to bring significant breakthroughs and application prospects to precise cancer medicine.

Abstract Rhodococcus equi (R. equi) is a zoonotic opportunistic pathogen that mainly causes fatal lung and extrapulmonary abscesses in foals and immunocompromised individuals. To date, no commercial vaccine against R. equi exists. We previously screened all potential vaccine candidates from the complete genome of R. equi using a reverse vaccinology approach. Five of these candidates, namely ABC transporter substrate-binding protein (ABC transporter), penicillin-binding protein 2 (PBD2), NlpC/P60 family protein (NlpC/P60), esterase family protein (Esterase), and M23 family metallopeptidase (M23) were selected for the evaluation of immunogenicity and immunoprotective effects in BALB/c mice model challenged with R. equi. The results showed that all five vaccine candidate-immunized mice experienced a significant increase in spleen antigen-specific IFN-γ- and TNF-α-positive CD4 + and CD8 + T lymphocytes and generated robust Th1- and Th2-type immune responses and antibody responses. Two weeks after the R. equi challenge, immunization with the five vaccine candidates reduced the bacterial load in the lungs and improved the pathological damage to the lungs and livers compared with those in the control group. NlpC/P60, Esterase, and M23 were more effective than the ABC transporter and PBD2 in inducing protective immunity against R. equi challenge in mice. In addition, these vaccine candidates have the potential to induce T lymphocyte memory immune responses in mice. In summary, these antigens are effective candidates for the development of protective vaccines against R. equi. The R. equi antigen library has been expanded and provides new ideas for the development of multivalent vaccines.

Coal fine intrusion into hydraulic propped fractures of coal seam easily leads to the blockage of the fracture, resulting in the decrease of the conductivity and the reduction of coal reservoir permeability, which seriously affects the stable discharge and production of coalbed methane well. In this paper, the permeability dynamic evolution models of coal fine intrusion into propped fractures were established before the shutdown and after the restart of coalbed methane well, and the experiments of coal fine intrusion into propped fracture under continuous and intermittent flow conditions were carried out by using the coal-rock conductivity test system, which verified the correctness of the models and studied the influence of coalbed methane well stoppage and drainage velocity on the permeability spatial and temporal evolution laws. The results indicate that with the increase of coal fine migration time, the permeability of propped fracture decreases slowly after a sudden drop, and along the direction of coal fine migration, the spatial pore loss rate of propped fracture decreases gradually, resulting in the permeability of fracture decreases along the direction of coal fine migration. With the continuous coal fine intrusion into propped fracture, the propped fracture pore loss rate after the shut-in and restart of coalbed methane wells is larger than that before the shutdown, and the permeability cannot be restored to the level before the shutdown. The slower the flow rate of drainage, the slower the permeability attenuation rate of coal fine intrusion into propped fracture, and the higher the permeability. During the coal fine invasion into proppant fracture at the low drainage flow rate, the fracture permeability is more sensitive to the flow rate change, and the less damage to the propped fracture permeability caused by the well shutdown. In the process of coal fine intrusion, the larger the deposition coefficient of coal fine is, the smaller the proximal fracture permeability and the larger the distal fracture permeability. The larger the diffusion coefficient is, the smaller the distal fracture permeability of the propped fracture is. The permeability of the proximal fracture is very little affected by the diffusion coefficient, and the damage to the proximal fracture permeability is more serious. In the process of coal fine intrusion, the proximal fracture permeability declines faster, while the distal fracture permeability declines slowly. The deposition coefficient changes have a significant influence on the proximal fracture permeability, while the diffusion coefficient changes have a more significant influence on the distal fracture permeability.