Graph Few-Shot Learning via Knowledge Transfer

Towards the challenging problem of semi-supervised node classification, there have been extensive studies. As a frontier, Graph Neural Networks (GNNs) have aroused great interest recently, which update the representation of each node by aggregating information of its neighbors. However, most GNNs have shallow layers with a limited receptive field and may not achieve satisfactory performance especially when the number of labeled nodes is quite small. To address this challenge, we innovatively propose a graph few-shot learning (GFL) algorithm that incorporates prior knowledge learned from auxiliary graphs to improve classification accuracy on the target graph. Specifically, a transferable metric space characterized by a node embedding and a graph-specific prototype embedding function is shared between auxiliary graphs and the target, facilitating the transfer of structural knowledge. Extensive experiments and ablation studies on four real-world graph datasets demonstrate the effectiveness of our proposed model and the contribution of each component.

Few-Shot Knowledge Graph Completion

Knowledge graphs (KGs) serve as useful resources for various natural language processing applications. Previous KG completion approaches require a large number of training instances (i.e., head-tail entity pairs) for every relation. The real case is that for most of the relations, very few entity pairs are available. Existing work of one-shot learning limits method generalizability for few-shot scenarios and does not fully use the supervisory information; however, few-shot KG completion has not been well studied yet. In this work, we propose a novel few-shot relation learning model (FSRL) that aims at discovering facts of new relations with few-shot references. FSRL can effectively capture knowledge from heterogeneous graph structure, aggregate representations of few-shot references, and match similar entity pairs of reference set for every relation. Extensive experiments on two public datasets demonstrate that FSRL outperforms the state-of-the-art.

Intra-articular injection of magnesium chloride attenuates osteoarthritis progression in rats

OBJECTIVE

To explore the effects of Mg²⁺ on the expression of osteoarthritic markers in human cartilage and synovium tissue explants. To investigate the therapeutic effect of intra-articular injection of Mg²⁺ in an established rat OA (Osteoarthritis) model of anterior cruciate ligament transection with partial medial meniscectomy (ACLT + PMM).

DESIGN

Human cartilage and synovium explants were collected from total knee replacement surgeries and incubated with MgCl₂ (20 mmol/L) in vitro. A rat OA model was established by ACLT + PMM surgery in 450-500 g male Sprague Dawley (SD) rats. To select the optimal dose, intra-articular injections of MgCl₂ (0.05, 0.5, 5 mol/L) were performed at 4 weeks after the surgery every 3 days for 2 weeks. The effect of optimized MgCl₂ was further determined by histology, immunohistochemistry, and quantitative real-time polymerase chain reaction.

RESULTS

The expressions of osteoarthritic markers in human cartilage and synovium explants were inhibited by Mg²⁺in vitro. Immunohistochemical analysis further suggested the inhibitory effects of Mg²⁺ on the expression of MMP-13 and IL-6 in the human tissue explants. Cartilage degeneration and synovitis in ACLT + PMM rats were significantly improved by intra-articular injections of Mg²⁺ (0.5 mol/L). Immunohistochemical analysis also showed the regulatory effects of Mg²⁺ on osteoarthritic markers in both cartilage and synovium in rats, consistent with in vitro results.

CONCLUSION

Intra-articular injections of Mg²⁺ at 0.5 mol/L attenuate the progression of OA in the ACLT + PMM rat model. Such effect was at least in part explained by the promotion of cartilage matrix synthesis and the suppression of synovial inflammation.

The role of miR-146b-5p in TLR4 pathway of glomerular mesangial cells with lupus nephritis

OBJECTIVE

MicroRNAs (miRNAs) are now recognized as important regulators of gene expression. The aim of the study was to investigate the role of miR-146b-5p in the TLR4 pathway and provide the basis for the treatment of lupus nephritis.

MATERIALS AND METHODS

The glomerular mesangial cells were cultured in vitro and divided into 3 groups: control group, a group of miR-146b-5p mimic added, and a group of miR-146b-5p inhibitor added. The levels of IL-6 and IL-8 in the cell culture supernatant of the three groups were detected by ELISA. The cell proliferation was detected by MTT. The expressions of MiR-146b-5p and TLR4 pathway-associated factor TRAF6 were detected by RT- PCR. The expression of TRAF6 and IRAK1 protein was detected by Western blot.

RESULTS

The overexpression of miR-146b-5p could reduce the level of IL6 and IL8 in cell culture and inhibit glomerular mesangial cell proliferation in some degree. Also, the overexpression of miR-146b-5p could inhibit the expressions of TLR4 pathway-associated factor TRAF6 and IRAK1mRNA, and the expressions of TRAF6 and IRAK1 protein.

CONCLUSIONS

MiR-146b-5p attenuated the inflammatory response of glomerular mesangial cells by inhibiting the expressions levels of TRAF6 and IRAK1 in lupus nephritis.

Baicalein, an enteric microbial metabolite, suppresses gut inflammation and cancer progression in ApcMin/+ mice

OBJECTIVE

Chronic inflammation is recognized as a risk factor for colorectal cancer (CRC) development. Baicalin (BI), a major constituent in an anti-inflammatory herb Scutellaria baicalensis, can be biotransformed into baicalein (BE) by the intestinal microbiota. We evaluated the anti-inflammation and anti-CRC effects of the metabolite BE.

METHODS

The in vitro biotransformation by human intestinal microbiota from BI into BE has been determined with HPLC. Using a gut-specific Apc^Min/+ mouse model, the effects of oral BE on the life span, organ index, and tumor multiplicity were evaluated. The expressions of inflammatory cytokines were determined using ELISA. To verify the in vivo data, the anti-inflammatory and antiproliferative effects of BE were determined with an in vitro cell model.

RESULTS

HPLC analysis showed that BI was quickly transformed into BE by the intestinal microbiota. Oral BE (30 mg/kg/day) significantly increased the life span, from 125.2 to 218.4 days (P < 0.01%). BE treatment also decreased intestine index and increased spleen index. Compared with the model group, following BE treatment, tumor numbers were significantly reduced in the small intestine and colon (P < 0.01, P < 0.05, respectively). In the gut tissues, BE treatment significantly reduced inflammatory cytokine levels such as IL-1β, IL-2, IL-6, IL-10, G-CSF, and GM-CSF. In vitro data supported our in vivo results that the anti-CRC effects of BE were via the inhibition of gut inflammation and induction of cancer cell death.

CONCLUSION

Our results suggest that the parent compound BI can be quickly converted into its microbial metabolite BE, which has stronger bioactive effects than BI. Baicalein is an active chemopreventive metabolite for inflammatory associated CRC.

2214Prevalence and severity of coronary disease in patients with familial hypercholesterolemia hospitalized for an acute myocardial infarction: data from the RICO survey

Aim

Individuals with heterozygous familial hypercholesterolemia (FH) are at high risk of early myocardial infarction (MI). However, coronary artery disease (CAD) burden of FH remains not well described. From a large database of a regional registry of acute MI, we aimed to address prevalence of FH and severity of CAD.

Methods

Consecutive patients hospitalized with MI in a multicentre database from 2001–2017 were considered. An algorithm, adapted from Dutch Lipid Clinic Network criteria, was built upon 4 variables (LDL-cholesterol (LDL-C) and lipid lowering agents, premature and family history of CAD) to identify FH probabilities.

Results

Among the 11624 patients included in the survey, 249 (2.1%) had probable/definite FH (score ≥6), and 2405 (20.7%) had possible FH (score 3–5). When compared with patients without FH (score 0–2), FH patients (score ≥6) were 20y younger (51 (46–57) vs 71 (61–80) y, p<0.001), with a lower rate of hypertension (47 vs 59%, p<0.001), diabetes (17 vs 25%, p<0.001) and prior stroke (4 vs 8%, p=0.016), but a higher prevalence of smokers (56 vs 23%, p<0.001), personal (20 vs 15%, p=0.02) or familial history of CAD (78 vs 18%, p<0.001). Chronic statin treatment was only used in 48% of FH patients and ezetimibe in 8%. After adjustment for age, sex and diabetes, FH patients were characterized by increased extent of CAD (syntax score 11 (4–19) vs 7 (1–13), p<0.001) and multivessel disease (55 vs 40%, p<0.001).

Conclusion

In this large real world population of acute MI, a high prevalence of FH was found. FH patients were characterized by their young age associated with the severity of CAD burden and limited use of preventive lipid lowering therapy.

Acknowledgement/Funding

University Hospital Center Dijon Bourgogne, Agence Régionale de Santé Bourgogne Franche Comté, France

Revisiting Spatial-Temporal Similarity: A Deep Learning Framework for Traffic Prediction

Traffic prediction has drawn increasing attention in AI research field due to the increasing availability of large-scale traffic data and its importance in the real world. For example, an accurate taxi demand prediction can assist taxi companies in pre-allocating taxis. The key challenge of traffic prediction lies in how to model the complex spatial dependencies and temporal dynamics. Although both factors have been considered in modeling, existing works make strong assumptions about spatial dependence and temporal dynamics, i.e., spatial dependence is stationary in time, and temporal dynamics is strictly periodical. However, in practice the spatial dependence could be dynamic (i.e., changing from time to time), and the temporal dynamics could have some perturbation from one period to another period. In this paper, we make two important observations: (1) the spatial dependencies between locations are dynamic; and (2) the temporal dependency follows daily and weekly pattern but it is not strictly periodic for its dynamic temporal shifting. To address these two issues, we propose a novel Spatial-Temporal Dynamic Network (STDN), in which a flow gating mechanism is introduced to learn the dynamic similarity between locations, and a periodically shifted attention mechanism is designed to handle long-term periodic temporal shifting. To the best of our knowledge, this is the first work that tackle both issues in a unified framework. Our experimental results on real-world traffic datasets verify the effectiveness of the proposed method.

Non-Stationary Model for Crime Rate Inference Using Modern Urban Data

Crime is one of the most important social problems in the country, affecting public safety, children development, and adult socioeconomic status. Understanding what factors cause higher crime rate is critical for policy makers in their efforts to reduce crime and increase citizens' life quality. We tackle a fundamental problem in our paper: crime rate inference at the neighborhood level. Traditional approaches have used demographics and geographical influences to estimate crime rates in a region. With the fast development of positioning technology and prevalence of mobile devices, a large amount of modern urban data have been collected and such big data can provide new perspectives for understanding crime. In this paper, we use large-scale Point-Of-Interest data and taxi flow data in the city of Chicago, IL in the USA. We observe significantly improved performance in crime rate inference compared to using traditional features. Such an improvement is consistent over multiple years. We also show that these new features are significant in the feature importance analysis. The correlations between crime and various observed features are not constant over the whole city. In order to address this geospatial non-stationary property, we further employ the geographically weighted regression on top of negative binomial model (GWNBR). Experiments have shown that GWNBR outperforms the negative binomial model.

Predicting Academic Performance for College Students

Detecting abnormal behaviors of students in time and providing personalized intervention and guidance at the early stage is important in educational management. Academic performance prediction is an important building block to enabling this pre-intervention and guidance. Most of the previous studies are based on questionnaire surveys and self-reports, which suffer from small sample size and social desirability bias. In this article, we collect longitudinal behavioral data from the smart cards of 6,597 students and propose three major types of discriminative behavioral factors, diligence, orderliness, and sleep patterns. Empirical analysis demonstrates these behavioral factors are strongly correlated with academic performance. Furthermore, motivated by the social influence theory, we analyze the correlation between each student’s academic performance with his/her behaviorally similar students’. Statistical tests indicate this correlation is significant. Based on these factors, we further build a multi-task predictive framework based on a learning-to-rank algorithm for academic performance prediction. This framework captures inter-semester correlation, inter-major correlation, and integrates student similarity to predict students’ academic performance. The experiments on a large-scale real-world dataset show the effectiveness of our methods for predicting academic performance and the effectiveness of proposed behavioral factors.

Effect of Sustained Joint Loading on TMJ Disc Nutrient Environment

The temporomandibular joint (TMJ) disc nutrient environment profoundly affects cell energy metabolism, proliferation, and biosynthesis. Due to technical challenges of in vivo measurements, the human TMJ disc extracellular nutrient environment under load, which depends on metabolic rates, solute diffusion, and disc morphometry, remains unknown. Therefore, the study objective was to predict the TMJ disc nutrient environment under loading conditions using combined experimental and computational modeling approaches. Specifically, glucose consumption and lactate production rates of porcine TMJ discs were measured under varying tissue culture conditions (n = 40 discs), and mechanical strain-dependent glucose and lactate diffusivities were measured using a custom diffusion chamber (n = 6 discs). TMJ anatomy and loading area were obtained from magnetic resonance imaging of healthy human volunteers (n = 11, male, 30 ± 9 y). Using experimentally determined nutrient metabolic rates, solute diffusivities, TMJ anatomy, and loading areas, subject-specific finite element (FE) models were developed to predict the 3-dimensional nutrient profiles in unloaded and loaded TMJ discs (unloaded, 0% strain, 20% strain). From the FE models, glucose, lactate, and oxygen concentration ranges for unloaded healthy human TMJ discs were 0.6 to 4.0 mM, 0.9 to 5.0 mM, and 0% to 6%, respectively, with steep gradients in the anterior and posterior bands. Sustained mechanical loading significantly reduced nutrient levels (P < 0.001), with a critical zone in which cells may die representing approximately 13.5% of the total disc volume. In conclusion, this study experimentally determined TMJ disc metabolic rates, solute diffusivities, and disc morphometry, and through subject-specific FE modeling, revealed critical interactions between mechanical loading and nutrient supply and metabolism for the in vivo human TMJ disc. The results suggest that TMJ disc homeostasis may be vulnerable to pathological loading (e.g., clenching, bruxism), which impedes nutrient supply. Given difficulties associated with direct in vivo measurements, this study provides a new approach to systematically investigate homeostatic and degenerative mechanisms associated with the TMJ disc.

The applied anatomy and clinical significance of the proximal, V1 segment of vertebral artery

BACKGROUND

The aim of the study was to probe the morphological features of the proximal segment (V1) of vertebral artery (VA) in a sample of Chinese cadavers.

MATERIALS AND METHODS

The origin, course and outer diameter at origin of the pre-vertebral part of the VAs were evaluated in 119 adult cadavers.

RESULTS

It was found that 94.12% of the VAs originated from the subclavian arteries, bilaterally. The variant origins were present in 5.88% of the cadavers and all originated directly from the arch of the aorta. All the variations were observed on the left side of male cadavers. The average outer diameters at origin of the normal and variation groups were 4.35 ± 1.00 mm and 4.82 ± ± 1.42 mm, respectively, p = 0.035. In the normal group, but not in the variation group, the average diameter in the males was significantly larger than that in the females (4.50 ± 0.99 mm, 3.92 ± 0.92 mm, respectively, p = 0.000). In addition, only 5 cadavers in the normal group had hypoplastic VAs (4.20%, 4 males, 3 right-sided). Vertebral artery dominance (VAD) was present in 91 (69 males) out of 112 cadavers and more common on the left (n = 48). In addition, 3 cadavers satisfied conditions for coexistence of VAD and vertebral artery hypoplasia. All 7 cadavers in the variation group exhibited VAD, which was more common on the right side (n = 5).

CONCLUSIONS

The morphologic variations and frequencies described above have implications for the early prevention, abnormal anatomy detection, accurate diagnosis, safe surgery and endovascular treatment of cardiovascular and neurological disease.

[Study on the correlation between work-related musculoskeletal disorders and work ability among nurses at three level general hospital in Xinjiang]

Objective: To investigate the correlation between work-related musculoskeletal disorders (WMSDs) and work ability (WA) among nurses in Xinjiang. Methods: A total of 1 873 nursing staff in the Departments of Internal Medicine, Surgery, Emergency and ICU Department, Operating Room and Supply Room of 6 Grade 3 General Hospitals in Xinjiang Beijiang Region were randomly selected by means of random sampling, and revised Nordic National Musculoskeletal Disorders Standard Questionnaire and Work Ability Index Scale were used to evaluate the status of nursing WMSDs and WA. Results: (1) The symptom incidence of work-related musculoskeletal disorders among nursing staff in the past year was 77.42%; The highest symptom incidence in all parts of the body was waist, neck, shoulder and back, and the symptom incidence was 63.59%, 59.80%, 47.78% and 37.32%, respectively; (2) The total score of nursing staff's work ability was (35.91±5.143) ; The proportion of "poor" , "middle" , "good" and "excellent" were 5.23%, 47.41%, 41.43% and 5.93% respectively; (3) The total working capacity of the sick group was (35.11±5.044) , The working ability of the non-diseased group was (38.64±4.507) , The difference between the two groups was statistically significant (t=13.773, P<0.01) ; There was a significant difference in the level of work ability between the two groups (z=-12.335, P<0.01) ; The working ability of "poor" and medium in the injury group was significantly higher than that of the non injury group, while the working ability was "better" and "excellent" than that of the non injury group; (4) The working ability of nursing staff of different age in WMSDs injury group was lower than that in non injury group, the difference was statistically significant (P<0.01) ; There was a negative correlation between the working ability of the injury group and the length of service, that is, with the increase of the length of service, the working ability index decreased (P<0.01) ; There was a positive correlation between the working ability of the non injury group and the length of service, that is, with the increase of the length of service, the work ability index increased gradually, but the correlation was not statistically significant (r=0.063, P=0.195) ; (5) The working ability of nursing staff of different departments in WMSDs injury group was lower than that in non injury group, the difference was statistically significant (P<0.01) . (6) Multiple linear regression analysis showed that the factors influencing the working ability of nursing staff were: past medical history, WMSDs disease, length of service and night shift frequency; Conclusion: The symptom incidence of WMSDs was high among nurses; Work Ability at a moderate level; work-related musculoskeletal disorders is associated with reduced ability of nursing staff; work-related musculoskeletal disorders are associated with reduced working capacity of nursing staff.

Abstract P4-04-11: A small amount of primary breast cancer shows high tumor mutation burden that may benefit from immune checkpoint inhibitor therapy

Background: Targeted therapy using immune checkpoint inhibitors (ICIs) is a major breakthrough in cancer treatment in the last decade. ICIs like PD1 or PD-L1 antibodies have been shown to be quite effective in cancer like melanoma. However, in most other tumor types including breast cancer, the situation is not as optimistic. Only a small percentage of those patients respond to ICIs therapy. This highlights the importance of identifying biomarkers to predict which patients may benefit from such treatment. Tumor Mutation Burden (TMB) has been shown to be a sensitive marker for ICI treatment. This study is to investigate whether TBM could be used as a biomarker for breast cancer treatment.

Methods: We reviewed next generation sequencing studies of breast cancer. Two such studies with raw data provided were included in our analysis. One study entitled METABRIC performed targeted sequencing of 173 cancer-related genes in around 2500 primary breast cancer tissues. The other study was from TCGA breast cancer project, which performed Whole Exome Sequencing (WES) of around 1000 primary breast cancer samples. Mutation data were downloaded from public data deposit. The number of mutations per sample was calculated. TBM was calculated by divide the coverage in million base pair from that of the total mutation counts.

Results: In METABRIC study, 17272 mutations were identified in 2369 samples, with a median of 7 mutations per sample (95% CI: 6 ˜ 7). The median TMB of METABRIC dataset was 5.8 SNVs/Mb (95% CI: 5 ˜ 5.8). Totally 30 out 2369 (1.3%) samples had a TMB equal or large than 20 SNVs/Mb. In another cohort from TCGA breast cancer study using WES technology, 90172 mutations were identified in 977 samples, with a median of 44 mutations per sample (95% CI: 39 ˜ 50). The median TMB was 1 SNVs/Mb (95% CI: 0.9 ˜ 1.1). Totally 13 out 977 (1.3%) samples had a TMB equal or large than 20 SNVs/Mb.

Conclusions: Breast cancer shows middle to low mutation burden compared to other cancer types. Around 1.3% of breast cancer has quite high TMB of at least 20 SNVs/Mb, which may be qualified for immune checkpoint inhibitors therapy. Our study indicates that TMB may be incorporated as a standard test for late stage breast cancer patients in the clinical practice.

Keywords: Breast cancer, Tumor Mutation Burden, Whole exome sequencing, Targeted sequencing, Immune checkpoint

Citation Format: Wang J, Chen W, Jiang Z, Lin X, Qin T, Yang X, Liu T, Hu H, Li Z, Xie D, Yao H, Song E. A small amount of primary breast cancer shows high tumor mutation burden that may benefit from immune checkpoint inhibitor therapy [abstract]. In: Proceedings of the 2018 San Antonio Breast Cancer Symposium; 2018 Dec 4-8; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2019;79(4 Suppl):Abstract nr P4-04-11.

Abstract P4-08-32: Derivation and validation of a novel prediction model in breast phyllodes tumors after surgery

Aim

This study aimed to develop a nomogram based on clinicopathological features to evaluate the recurrence probability of breast phyllodes tumors following surgery. The criteria for atypia, mitoses, overgrowth, and surgical margin (AMOS) were also validated.

Method

Data from 334 patients with breast phyllodes tumors, who underwent surgical treatment at Sun Yat-sen Memorial Hospital from January 2005 to December 2014, were used to develop a prediction model. Additionally, data of 36 patients from Peking University Shenzhen Hospital and data of 140 patients from Sun Yat-sen University Cancer Center during the same period were used to validate the model. The medical records and tumor slides were retrospectively reviewed. The log-rank and Cox regression tests were used to develop a clinical prediction model of breast phyllodes tumors as well as validating the AMOS criteriaAll statistical analyses were performed using R and STATA.

Results

Of all 334 patients included in the study, 224 had benign, 91 had borderline, and 19 had malignant tumors. The local and distant recurrence rate was 17.7%. The 1-,3-, and 5-year cumulative recurrence-free survival was 98.5%, 97.9%, and 96.8%, respectively. Surgical margin, mitoses, and tumor border were identified as independent risk factors for breast phyllodes tumors. A nomogram was developed based on these three variables. The C-index of internal and external validation was 0.71and 0.67, respectively. The area under the curve of AMOS criteria was 0.59.

Conclusions

The present study model presented a more concise and objective variables to evaluate the recurrence-free survival of patients after surgery compared with that using the AMOS criteria, which is more appropriate for clinical practice and also allows for a more accurate prediction.

Citation Format: Song E, Chao X, Nie Y, Jin X, Tan C, Cui J, Hu H, Yao H. Derivation and validation of a novel prediction model in breast phyllodes tumors after surgery [abstract]. In: Proceedings of the 2018 San Antonio Breast Cancer Symposium; 2018 Dec 4-8; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2019;79(4 Suppl):Abstract nr P4-08-32.