Outcome Prediction Using Multi-Modal Information: Integrating Large Language Model-Extracted Clinical Information and Image Analysis.

Survival prediction post-cystectomy is essential for the follow-up care of bladder cancer patients. This study aimed to evaluate artificial intelligence (AI)-large language models (LLMs) for extracting clinical information and improving image analysis, with an initial application involving predicting five-year survival rates of patients after radical cystectomy for bladder cancer. Data were retrospectively collected from medical records and CT urograms (CTUs) of bladder cancer patients between 2001 and 2020. Of 781 patients, 163 underwent chemotherapy, had pre- and post-chemotherapy CTUs, underwent radical cystectomy, and had an available post-surgery five-year survival follow-up. Five AI-LLMs (Dolly-v2, Vicuna-13b, Llama-2.0-13b, GPT-3.5, and GPT-4.0) were used to extract clinical descriptors from each patient's medical records. As a reference standard, clinical descriptors were also extracted manually. Radiomics and deep learning descriptors were extracted from CTU images. The developed multi-modal predictive model, CRD, was based on the clinical (C), radiomics (R), and deep learning (D) descriptors. The LLM retrieval accuracy was assessed. The performances of the survival predictive models were evaluated using AUC and Kaplan-Meier analysis. For the 163 patients (mean age 64 ± 9 years; M:F 131:32), the LLMs achieved extraction accuracies of 74%~87% (Dolly), 76%~83% (Vicuna), 82%~93% (Llama), 85%~91% (GPT-3.5), and 94%~97% (GPT-4.0). For a test dataset of 64 patients, the CRD model achieved AUCs of 0.89 ± 0.04 (manually extracted information), 0.87 ± 0.05 (Dolly), 0.83 ± 0.06~0.84 ± 0.05 (Vicuna), 0.81 ± 0.06~0.86 ± 0.05 (Llama), 0.85 ± 0.05~0.88 ± 0.05 (GPT-3.5), and 0.87 ± 0.05~0.88 ± 0.05 (GPT-4.0). This study demonstrates the use of LLM model-extracted clinical information, in conjunction with imaging analysis, to improve the prediction of clinical outcomes, with bladder cancer as an initial example.

Flow2GNN: Flexible Two-Way Flow Message Passing for Enhancing GNNs Beyond Homophily.

Message passing (MP) is crucial for effective graph neural networks (GNNs). Most local message-passing schemes have been shown to underperform on heterophily graphs due to the perturbation of updated representations caused by local redundant heterophily information. However, our experiment findings indicate that the distribution of heterophily information during MP can be disrupted by disentangling local neighborhoods. This finding can be applied to other GNNs, improving their performance on heterophily graphs in a more flexible manner compared to most heterophily GNNs with complex designs. This article proposes a new type of simple message-passing neural network called Flow2GNN. It uses a two-way flow message-passing scheme to enhance the ability of GNNs by disentangling and redistributing heterophily information in the topology space and the attribute space. Our proposed message-passing scheme consists of two steps in topology space and attribute space. First, we introduce a new disentangled operator with binary elements that disentangle topology information in-flow and out-flow between connected nodes. Second, we use an adaptive aggregation model that adjusts the flow amount between homophily and heterophily attribute information. Furthermore, we rigorously prove that disentangling in message-passing can reduce the generalization gap, offering a deeper understanding of how our model enhances other GNNs. The extensive experiment results show that the proposed model, Flow2GNN, not only outperforms state-of-the-art GNNs, but also helps improve the performance of other commonly used GNNs on heterophily graphs, including GCN, GAT, GCNII, and H 2 GCN, specifically for GCN, with up to a 25.88% improvement on the Wisconsin dataset.

Enhancing Early Lung Cancer Diagnosis: Predicting Lung Nodule Progression in Follow-Up Low-Dose CT Scan with Deep Generative Model.

Early diagnosis of lung cancer can significantly improve patient outcomes. We developed a Growth Predictive model based on the Wasserstein Generative Adversarial Network framework (GP-WGAN) to predict the nodule growth patterns in the follow-up LDCT scans. The GP-WGAN was trained with a training set (N = 776) containing 1121 pairs of nodule images with about 1-year intervals and deployed to an independent test set of 450 nodules on baseline LDCT scans to predict nodule images (GP-nodules) in their 1-year follow-up scans. The 450 GP-nodules were finally classified as malignant or benign by a lung cancer risk prediction (LCRP) model, achieving a test AUC of 0.827 ± 0.028, which was comparable to the AUC of 0.862 ± 0.028 achieved by the same LCRP model classifying real follow-up nodule images (p = 0.071). The net reclassification index yielded consistent outcomes (NRI = 0.04; p = 0.62). Other baseline methods, including Lung-RADS and the Brock model, achieved significantly lower performance (p < 0.05). The results demonstrated that the GP-nodules predicted by our GP-WGAN model achieved comparable performance with the nodules in the real follow-up scans for lung cancer diagnosis, indicating the potential to detect lung cancer earlier when coupled with accelerated clinical management versus the current approach of waiting until the next screening exam.

Adolescent predictors of young adult pain and health outcomes: results from a 6-year prospective follow-up study.

ABSTRACT: Adolescent chronic pain may lead to persistent disability and long-term health impairments in adulthood. However, our understanding of which youth are more likely to experience adverse outcomes remains limited. To address this gap, this longitudinal cohort study examined adolescent predictors of various dimensions of young adult health and functioning, including pain, physical health, depression, anxiety, social isolation, and sleep disturbance. As part of a previous clinical trial, we recruited a cohort of adolescents (ages 11-17 years, M age = 14 years) with non-disease-related chronic pain from 15 tertiary pain clinics in North America. Approximately 6 years later, 229 of the original 273 individuals (81% participation rate) completed a follow-up survey as young adults (ages 18-25 years, M age = 21 years). At the young adult follow-up, 73% reported continued chronic pain, with two-thirds experiencing moderate-to-severe pain interference. Youth reported several adverse health outcomes, including below-average physical health (37%), clinically elevated depression (42%), clinically elevated anxiety (48%), and sleep disturbances (77%). Multivariate regression analyses controlling for sociodemographic characteristics revealed that higher pain intensity, more pain locations, lower sleep quality, and greater anxiety symptoms in adolescence predicted worse pain outcomes in young adulthood. Moreover, lower sleep quality, greater anxiety symptoms, and worse family functioning predicted worse physical and psychosocial health in adulthood. These findings represent an important first step toward identifying ways to optimize psychological pain interventions. Tailored psychological pain interventions can directly target adolescent vulnerabilities, including mood, sleep, and family risk factors, with the potential to disrupt a lifelong trajectory of pain and suffering.

TUFM in health and disease: exploring its multifaceted roles.

The nuclear-encoded mitochondrial protein Tu translation elongation factor, mitochondrial (TUFM) is well-known for its role in mitochondrial protein translation. Originally discovered in yeast, TUFM demonstrates significant evolutionary conservation from prokaryotes to eukaryotes. Dysregulation of TUFM has been associated with mitochondrial disorders. Although early hypothesis suggests that TUFM is localized within mitochondria, recent studies identify its presence in the cytoplasm, with this subcellular distribution being linked to distinct functions of TUFM. Significantly, in addition to its established function in mitochondrial protein quality control, recent research indicates a broader involvement of TUFM in the regulation of programmed cell death processes (e.g., autophagy, apoptosis, necroptosis, and pyroptosis) and its diverse roles in viral infection, cancer, and other disease conditions. This review seeks to offer a current summary of TUFM's biological functions and its complex regulatory mechanisms in human health and disease. Insight into these intricate pathways controlled by TUFM may lead to the potential development of targeted therapies for a range of human diseases.

Bunyavirus SFTSV NSs utilizes autophagy to escape the antiviral innate immune response.

Severe fever with thrombocytopenia syndrome virus (SFTSV) nonstructural protein (NSs) is an important viral virulence factor that sequesters multiple antiviral proteins into inclusion bodies to escape the antiviral innate immune response. However, the mechanism of the NSs restricting host innate immunity remains largely elusive. Here, we found that the NSs induced complete macroautophagy/autophagy by interacting with the CCD domain of BECN1, thereby promoting the formation of a BECN1-dependent autophagy initiation complex. Importantly, our data showed that the NSs sequestered antiviral proteins such as TBK1 into autophagic vesicles, and therefore promoted the degradation of TBK1 and other antiviral proteins. In addition, the 8A mutant of NSs reduced the induction of BECN1-dependent autophagy flux and degradation of antiviral immune proteins. In conclusion, our results indicated that SFTSV NSs sequesters antiviral proteins into autophagic vesicles for degradation and to escape antiviral immune responses.

Leveraging Serial Low-Dose CT Scans in Radiomics-based Reinforcement Learning to Improve Early Diagnosis of Lung Cancer at Baseline Screening.

Purpose To evaluate the feasibility of leveraging serial low-dose CT (LDCT) scans to develop a radiomics-based reinforcement learning (RRL) model for improving early diagnosis of lung cancer at baseline screening. Materials and Methods In this retrospective study, 1951 participants (female patients, 822; median age, 61 years [range, 55-74 years]) (male patients, 1129; median age, 62 years [range, 55-74 years]) were randomly selected from the National Lung Screening Trial between August 2002 and April 2004. An RRL model using serial LDCT scans (S-RRL) was trained and validated using data from 1404 participants (372 with lung cancer) containing 2525 available serial LDCT scans up to 3 years. A baseline RRL (B-RRL) model was trained with only LDCT scans acquired at baseline screening for comparison. The 547 held-out individuals (150 with lung cancer) were used as an independent test set for performance evaluation. The area under the receiver operating characteristic curve (AUC) and the net reclassification index (NRI) were used to assess the performances of the models in the classification of screen-detected nodules. Results Deployment to the held-out baseline scans showed that the S-RRL model achieved a significantly higher test AUC (0.88 [95% CI: 0.85, 0.91]) than both the Brock model (AUC, 0.84 [95% CI: 0.81, 0.88]; P = .02) and the B-RRL model (AUC, 0.86 [95% CI: 0.83, 0.90]; P = .02). Lung cancer risk stratification was significantly improved by the S-RRL model as compared with Lung CT Screening Reporting and Data System (NRI, 0.29; P < .001) and the Brock model (NRI, 0.12; P = .008). Conclusion The S-RRL model demonstrated the potential to improve early diagnosis and risk stratification for lung cancer at baseline screening as compared with the B-RRL model and clinical models. Keywords: Radiomics-based Reinforcement Learning, Lung Cancer Screening, Low-Dose CT, Machine Learning © RSNA, 2024 Supplemental material is available for this article.

SFTSV nucleoprotein mediates DNA sensor cGAS degradation to suppress cGAS-dependent antiviral responses.

Cyclic GMP-AMP synthase (cGAS) is an important DNA pattern recognition receptor that senses double-stranded DNA derived from invading pathogens or self DNA in cytoplasm, leading to an antiviral interferon response. A tick-borne Bunyavirus, severe fever with thrombocytopenia syndrome virus (SFTSV), is an RNA virus that causes a severe emerging viral hemorrhagic fever in Asia with a high case fatality rate of up to 30%. However, it is unclear whether cGAS interacts with SFTSV infection. In this study, we found that SFTSV infection upregulated cGAS RNA transcription and protein expression, indicating that cGAS is an important innate immune response against SFTSV infection. The mechanism of cGAS recognizing SFTSV is by cGAS interacting with misplaced mitochondrial DNA in the cytoplasm. Depletion of mitochondrial DNA significantly inhibited cGAS activation under SFTSV infection. Strikingly, we found that SFTSV nucleoprotein (N) induced cGAS degradation in a dose-dependent manner. Mechanically, N interacted with the 161-382 domain of cGAS and linked the cGAS to LC3. The cGAS-N-LC3 trimer was targeted to N-induced autophagy, and the cGAS was degraded in autolysosome. Taken together, our study discovered a novel antagonistic mechanism of RNA viruses, SFTSV is able to suppress the cGAS-dependent antiviral innate immune responses through N-hijacking cGAS into N-induced autophagy. Our results indicated that SFTSV N is an important virulence factor of SFTSV in mediating host antiviral immune responses. IMPORTANCE: Severe fever with thrombocytopenia syndrome virus (SFTSV) is a tick-borne RNA virus that is widespread in East and Southeast Asian countries with a high fatality rate of up to 30%. Up to now, many cytoplasmic pattern recognition receptors, such as RIG-I, MDA5, and SAFA, have been reported to recognize SFTSV genomic RNA and trigger interferon-dependent antiviral responses. However, current knowledge is not clear whether SFTSV can be recognized by DNA sensor cyclic GMP-AMP synthase (cGAS). Our study demonstrated that cGAS could recognize SFTSV infection via ectopic mitochondrial DNA, and the activated cGAS-stimulator of interferon genes signaling pathway could significantly inhibit SFTSV replication. Importantly, we further uncovered a novel mechanism of SFTSV to inhibit innate immune responses by the degradation of cGAS. cGAS was degraded in N-induced autophagy. Collectively, this study illustrated a novel virulence factor of SFTSV to suppress innate immune responses through autophagy-dependent cGAS degradation.

Antibiotic impregnated catheters and intrathecal antibiotics for CSF shunt infection prevention in children undergoing low-risk CSF shunt surgery.

BACKGROUND: Cerebrospinal fluid (CSF) shunts allow children with hydrocephalus to survive and avoid brain injury (J Neurosurg 107:345-57, 2007; Childs Nerv Syst 12:192-9, 1996). The Hydrocephalus Clinical Research Network implemented non-randomized quality improvement protocols that were shown to decrease infection rates compared to pre-operative prophylactic intravenous antibiotics alone (standard care): initially with intrathecal (IT) antibiotics between 2007-2009 (J Neurosurg Pediatr 8:22-9, 2011), followed by antibiotic impregnated catheters (AIC) in 2012-2013 (J Neurosurg Pediatr 17:391-6, 2016). No large scale studies have compared infection prevention between the techniques in children. Our objectives were to compare the risk of infection following the use of IT antibiotics, AIC, and standard care during low-risk CSF shunt surgery (i.e., initial CSF shunt placement and revisions) in children. METHODS: A retrospective observational cohort study at 6 tertiary care children's hospitals was conducted using Pediatric Health Information System + (PHIS +) data augmented with manual chart review. The study population included children ≤ 18 years who underwent initial shunt placement between 01/2007 and 12/2012. Infection and subsequent CSF shunt surgery data were collected through 12/2015. Propensity score adjustment for regression analysis was developed based on site, procedure type, and year; surgeon was treated as a random effect. RESULTS: A total of 1723 children underwent initial shunt placement between 2007-2012, with 1371 subsequent shunt revisions and 138 shunt infections. Propensity adjusted regression demonstrated no statistically significant difference in odds of shunt infection between IT antibiotics (OR 1.22, 95% CI 0.82-1.81, p = 0.3) and AICs (OR 0.91, 95% CI 0.56-1.49, p = 0.7) compared to standard care. CONCLUSION: In a large, observational multicenter cohort, IT antibiotics and AICs do not confer a statistically significant risk reduction compared to standard care for pediatric patients undergoing low-risk (i.e., initial or revision) shunt surgeries.

Evaluation of a walking school bus program: a cluster randomized controlled trial.

BACKGROUND: The purpose of this study was to investigate the effects of a walking school bus intervention on children's active commuting to school. METHODS: We conducted a randomized controlled trial (RCT) in Houston, Texas (Year 1) and Seattle, Washington (Years 2-4) from 2012 to 2016. The study had a two-arm, cluster randomized design comparing the intervention (walking school bus and education materials) to the control (education materials) over one school year October/November - May/June). Twenty-two schools that served lower income families participated. Outcomes included percentage of days students' active commuting to school (primary, measured via survey) and moderate-to-vigorous physical activity (MVPA, measured via accelerometry). Follow-up took place in May or June. We used linear mixed-effects models to estimate the association between the intervention and outcomes of interest. RESULTS: Total sample was 418 students [Mage=9.2 (SD = 0.9) years; 46% female], 197 (47%) in the intervention group. The intervention group showed a significant increase compared with the control group over time in percentage of days active commuting (ß = 9.04; 95% CI: 1.10, 16.98; p = 0.015) and MVPA minutes/day (ß = 4.31; 95% CI: 0.70, 7.91; p = 0.02). CONCLUSIONS: These findings support implementation of walking school bus programs that are inclusive of school-age children from lower income families to support active commuting to school and improve physical activity. TRAIL REGISTRATION: This RCT is registered at clinicaltrials.gov (NCT01626807).

Reducing IV Antibiotic Duration for Neonatal UTI Using a Clinical Standard Pathway.

OBJECTIVES: Urinary tract infections (UTIs) are the most common bacterial infections in young infants and are traditionally treated with longer intravenous (IV) antibiotic courses. A growing body of evidence supports shorter IV antibiotic courses for young infants. Our primary aim was to decrease the IV antibiotic treatment to 3 days over 2 years for neonates aged 0 to 28 days who have been hospitalized with UTIs. METHODS: Using quality improvement methods, our primary intervention was to implement a revised clinical pathway recommending 3 (previously 7) days of IV antibiotics. Our primary outcome measure was IV antibiotic duration, and the secondary outcomes were length of stay (LOS) and costs. The balancing measure was readmission within 30 days of discharge. Neonates were identified by using International Classification of Diseases diagnosis codes and excluded if they were admitted to the ICU or had a LOS >30 days. We used statistical process control to analyze outcome measures for 4 years before (baseline) and 2 years after the pathway revision (intervention) in February 2020. RESULTS: A total of 93 neonates were hospitalized with UTIs in the baseline period and 41 were hospitalized in the intervention period. We found special cause variation, with a significant decrease in mean IV antibiotic duration from 4.7 (baseline) to 3.1 days (intervention) and a decrease in mean LOS from 5.4 to 3.6 days. Costs did not differ between the baseline and intervention periods. There were 7 readmissions during the baseline period, and 0 during the intervention period. CONCLUSIONS: The implementation of a revised clinical pathway significantly reduced IV antibiotic treatment duration and hospital LOS for neonatal UTIs without an increase in hospital readmissions.

Design, synthesis, and bioevaluation of SOS1 PROTACs derived from pyrido[2,3-d]pyrimidin-7-one-based SOS1 inhibitor.

Oncogenic KRAS mutations drive an approximately 25 % of all human cancers. Son of Sevenless 1 (SOS1), a critical guanine nucleotide exchange factor, catalyzes the activation of KRAS. Targeting SOS1 degradation has engaged as a promising therapeutic strategy for KRAS-mutant cancers. Herein, we designed and synthesized a series of novel CRBN-recruiting SOS1 PROTACs using the pyrido[2,3-d]pyrimidin-7-one-based SOS1 inhibitor as the warhead. One representative compound 11o effectively induced the degradation of SOS1 in three different KRAS-mutant cancer cell lines with DC50 values ranging from 1.85 to 7.53 nM. Mechanism studies demonstrated that 11o-induced SOS1 degradation was dependent on CRBN and proteasome. Moreover, 11o inhibited the phosphorylation of ERK and displayed potent anti-proliferative activities against SW620, A549 and DLD-1 cells. Further optimization of 11o may provide us promising SOS1 degraders with favorable drug-like properties for developing new chemotherapies targeting KRAS-driven cancers.

Enrooted-Type Metal-Support Interaction Boosting Oxygen Evolution Reaction in Acidic Media.

Supported metal catalysts with appropriate metal-support interactions (MSIs) hold a great promise for heterogeneous catalysis. However, ensuring tight immobilization of metal clusters/nanoparticles on the support while maximizing the exposure of surface active sites remains a huge challenge. Herein, we report an Ir/WO3 catalyst with a new enrooted-type MSI in which Ir clusters are, unprecedentedly, atomically enrooted into the WO3 lattice. The enrooted Ir atoms decrease the electron density of the constructed interface compared to the adhered (root-free) type, thereby achieving appropriate adsorption toward oxygen intermediates, ultimately leading to high activity and stability for oxygen evolution in acidic media. Importantly, this work provides a new enrooted-type supported metal catalyst, which endows suitable MSI and maximizes the exposure of surface active sites in contrast to the conventional adhered, embedded, and encapsulated types.

PON3::LCN1 and HTN3::MSANTD3 Gene Fusions With NR4A3/NR4A2 Expression in Salivary Acinic Cell Carcinoma.

Acinic cell carcinoma of the salivary gland (AciCC) is a low-grade carcinoma characterized by the overexpression of the transcription factor nuclear receptor subfamily 4 group A member 3 (NR4A3). AciCC has been the subject of a few molecular research projects. This study delves into AciCC's molecular landscape to identify additional alterations and explore their clinical implications. RNA sequencing and immunohistochemical staining for markers NR4A3/NR4A2, DOG-1, S100, and mammaglobin were utilized on 41 AciCCs and 11 secretory carcinoma (SC) samples. NR4A3 was evident in 35 AciCCs, while the residual 6 were NR4A3-negative and NR4A2-positive; SC samples were consistently NR4A3-negative. A novel fusion, PON3 exon 1- LCN1 exon 5, was detected in 9/41 (21.9%) AciCCs, exhibiting a classical histologic pattern with serous cell components growing in solid sheets alongside the intercalated duct-like component. Clinical follow-up of 39 patients over a median of 59 months revealed diverse prognostic outcomes: 34 patients exhibited no disease evidence, whereas the remaining 5 experienced poorer prognosis, involving local recurrence, lymph node, and distant metastasis, and disease-associated death, 4 of which harbored the PON3::LCN1 fusion. In addition, the HTN3::MSANTD3 fusion was recurrently identified in 7/41 AciCC cases. SC patients lacked both fusions. Immunohistochemistry uncovered differential expression of DOG-1, S100, and mammaglobin across samples, providing nuanced insights into their roles in AciCC. This study accentuates PON3::LCN1 and HTN3::MSANTD3 fusions as recurrent molecular events in AciCC, offering potential diagnostic and prognostic utility and propelling further research into targeted therapeutic strategies.

Multimodal data integration for predicting progression risk in castration-resistant prostate cancer using deep learning: a multicenter retrospective study.

Purpose: Patients with advanced prostate cancer (PCa) often develop castration-resistant PCa (CRPC) with poor prognosis. Prognostic information obtained from multiparametric magnetic resonance imaging (mpMRI) and histopathology specimens can be effectively utilized through artificial intelligence (AI) techniques. The objective of this study is to construct an AI-based CRPC progress prediction model by integrating multimodal data. Methods and materials: Data from 399 patients diagnosed with PCa at three medical centers between January 2018 and January 2021 were collected retrospectively. We delineated regions of interest (ROIs) from 3 MRI sequences viz, T2WI, DWI, and ADC and utilized a cropping tool to extract the largest section of each ROI. We selected representative pathological hematoxylin and eosin (H&E) slides for deep-learning model training. A joint combined model nomogram was constructed. ROC curves and calibration curves were plotted to assess the predictive performance and goodness of fit of the model. We generated decision curve analysis (DCA) curves and Kaplan-Meier (KM) survival curves to evaluate the clinical net benefit of the model and its association with progression-free survival (PFS). Results: The AUC of the machine learning (ML) model was 0.755. The best deep learning (DL) model for radiomics and pathomics was the ResNet-50 model, with an AUC of 0.768 and 0.752, respectively. The nomogram graph showed that DL model contributed the most, and the AUC for the combined model was 0.86. The calibration curves and DCA indicate that the combined model had a good calibration ability and net clinical benefit. The KM curve indicated that the model integrating multimodal data can guide patient prognosis and management strategies. Conclusion: The integration of multimodal data effectively improves the prediction of risk for the progression of PCa to CRPC.

Retracted: Resveratrol Suppresses Human Nasopharyngeal Carcinoma Cell Growth Via Inhibiting Differentiation Antagonizing Non-Protein Coding RNA (DANCR) Expression.

The authors have requested retraction due to the identification of errors in the data. Reference: Jiafeng Zhang, Xiaojie Jin, Chuan Zhou, Hui Zhao, Ping He, Yalin Hao, Qiongna Dong. Resveratrol Suppresses Human Nasopharyngeal Carcinoma Cell Growth Via Inhibiting Differentiation Antagonizing Non-Protein Coding RNA (DANCR) Expression. Med Sci Monit, 2020; 26: e923622. DOI: 10.12659/MSM.923622.

Computed tomography-based nomogram of Siewert type II/III adenocarcinoma of esophagogastric junction to predict response to docetaxel, oxaliplatin and S-1.

BACKGROUND: Neoadjuvant chemotherapy (NAC) has become the standard care for advanced adenocarcinoma of esophagogastric junction (AEG), although a part of the patients cannot benefit from NAC. There are no models based on baseline computed tomography (CT) to predict response of Siewert type II or III AEG to NAC with docetaxel, oxaliplatin and S-1 (DOS). AIM: To develop a CT-based nomogram to predict response of Siewert type II/III AEG to NAC with DOS. METHODS: One hundred and twenty-eight consecutive patients with confirmed Siewert type II/III AEG underwent CT before and after three cycles of NAC with DOS, and were randomly and consecutively assigned to the training cohort (TC) (n = 94) and the validation cohort (VC) (n = 34). Therapeutic effect was assessed by disease-control rate and progressive disease according to the Response Evaluation Criteria in Solid Tumors (version 1.1) criteria. Possible prognostic factors associated with responses after DOS treatment including Siewert classification, gross tumor volume (GTV), and cT and cN stages were evaluated using pretherapeutic CT data in addition to sex and age. Univariate and multivariate analyses of CT and clinical features in the TC were performed to determine independent factors associated with response to DOS. A nomogram was established based on independent factors to predict the response. The predictive performance of the nomogram was evaluated by Concordance index (C-index), calibration and receiver operating characteristics curve in the TC and VC. RESULTS: Univariate analysis showed that Siewert type (52/55 vs 29/39, P = 0.005), pretherapeutic cT stage (57/62 vs 24/32, P = 0.028), GTV (47.3 ± 27.4 vs 73.2 ± 54.3, P = 0.040) were significantly associated with response to DOS in the TC. Multivariate analysis of the TC also showed that the pretherapeutic cT stage, GTV and Siewert type were independent predictive factors related to response to DOS (odds ratio = 4.631, 1.027 and 7.639, respectively; all P < 0.05). The nomogram developed with these independent factors showed an excellent performance to predict response to DOS in the TC and VC (C-index: 0.838 and 0.824), with area under the receiver operating characteristic curve of 0.838 and 0.824, respectively. The calibration curves showed that the practical and predicted response to DOS effectively coincided. CONCLUSION: A novel nomogram developed with pretherapeutic cT stage, GTV and Siewert type predicted the response of Siewert type II/III AEG to NAC with DOS.

Local-environment-guided selection of atomic structures for the development of machine-learning potentials.

Machine learning potentials (MLPs) have attracted significant attention in computational chemistry and materials science due to their high accuracy and computational efficiency. The proper selection of atomic structures is crucial for developing reliable MLPs. Insufficient or redundant atomic structures can impede the training process and potentially result in a poor quality MLP. Here, we propose a local-environment-guided screening algorithm for efficient dataset selection in MLP development. The algorithm utilizes a local environment bank to store unique local environments of atoms. The dissimilarity between a particular local environment and those stored in the bank is evaluated using the Euclidean distance. A new structure is selected only if its local environment is significantly different from those already present in the bank. Consequently, the bank is then updated with all the new local environments found in the selected structure. To demonstrate the effectiveness of our algorithm, we applied it to select structures for a Ge system and a Pd13H2 particle system. The algorithm reduced the training data size by around 80% for both without compromising the performance of the MLP models. We verified that the results were independent of the selection and ordering of the initial structures. We also compared the performance of our method with the farthest point sampling algorithm, and the results show that our algorithm is superior in both robustness and computational efficiency. Furthermore, the generated local environment bank can be continuously updated and can potentially serve as a growing database of feature local environments, aiding in efficient dataset maintenance for constructing accurate MLPs.

Knowledge Distillation Guided Interpretable Brain Subgraph Neural Networks for Brain Disorder Exploration.

The human brain is a highly complex neurological system that has been the subject of continuous exploration by scientists. With the help of modern neuroimaging techniques, there has been significant progress made in brain disorder analysis. There is an increasing interest about utilizing artificial intelligence techniques to improve the efficiency of disorder diagnosis in recent years. However, these methods rely only on neuroimaging data for disorder diagnosis and do not explore the pathogenic mechanism behind the disorder or provide an interpretable result toward the diagnosis decision. Furthermore, the scarcity of medical data limits the performance of existing methods. As the hot application of graph neural networks (GNNs) in molecular graphs and drug discovery due to its strong graph-structured data learning ability, whether GNNs can also play a huge role in the field of brain disorder analysis. Thus, in this work, we innovatively model brain neuroimaging data into graph-structured data and propose knowledge distillation (KD) guided brain subgraph neural networks to extract discriminative subgraphs between patient and healthy brain graphs to explain which brain regions and abnormal functional connectivities cause the disorder. Specifically, we introduce the KD technique to transfer the knowledge of pretrained teacher model to guide brain subgraph neural networks training and alleviate the problem of insufficient training data. And these discriminative subgraphs are conducive to learn better brain graph-level representations for disorder prediction. We conduct abundant experiments on two functional magnetic resonance imaging datasets, i.e., Parkinson's disease (PD) and attention-deficit/hyperactivity disorder (ADHD), and experimental results well demonstrate the superiority of our method over other brain graph analysis methods for disorder prediction accuracy. The interpretable experimental results given by our method are consistent with corresponding medical research, which is encouraging to provide a potential for deeper brain disorder study.