Search

As the training dataset size and the model size of machine learning increase rapidly, more computing resources are consumed to speedup the training process. However, the scalability and performance reproducibility of parallel machine learning training, which mainly uses stochastic optimization algorithms, are limited. In this paper, we demonstrate that the sample difference in the dataset plays a prominent role in the scalability of parallel machine learning algorithms. We propose to use statistical properties of dataset to measure sample differences. These properties include the variance of sample features, sample sparsity, sample diversity, and similarity in sampling sequences. We choose four types of parallel training algorithms as our research objects: (1) the asynchronous parallel SGD algorithm (Hogwild! algorithm), (2) the parallel model average SGD algorithm (minibatch SGD algorithm), (3) the decentralization optimization algorithm, and (4) the dual coordinate optimization (DADM algorithm). Our results show that the statistical properties of training datasets determine the scalability upper bound of these parallel training algorithms.

Poly(ADP-ribose) polymerase (PARP) inhibitors represent one of the successful novel approaches to targeted cancer treatment. Indeed, the US Food and Drug Administration (FDA) has recently approved PARP inhibitors for the treatment of breast and ovarian cancers. Despite the proven efficacy of these agents, certain challenges remain with their use. Among the most important are primary and secondary resistance. Here, we review the mechanism of action of PARP inhibitors and their ability to exploit certain inherent deficiencies among malignant cells to improve cell killing, with a focus on deficiencies in homologous recombination among cells with BRCA1 and BRCA2 mutations. Moreover, we discuss the different mechanisms of resistance including development of secondary resistance and strategies to overcome them. Finally, we discuss the limitations of novel therapeutic interventions and possible future studies to exploit biochemical pathways in order to improve therapeutic efficacy of PARP inhibitors.

AimPost-operative fatigue syndrome (POFS) is a common complication that prolongs the recovery to normal function and activity after surgery. The aim of the present study was to explore the mechanism of central fatigue in POFS and the anti-fatigue effect of ginsenoside Rb1.MethodWe investigated the association between inflammation, indoleamine 2,3-dioxygenase (IDO) enzyme, and tryptophan metabolism in the hippocampus of POFS rats. A POFS rat model was induced by major small intestinal resection. Rats with major small intestinal resection were administered ginsenoside Rb1 (15 mg/kg) once a day from 3 days before surgery to the day of sacrifice, or with saline as corresponding controls. Fatigue was assessed with the open field test (OFT) and sucrose preference test (SPT). ELISA, RT-PCR, Western blot, immunofluorescence, and high-performance liquid chromatography (HPLC) were used to test the inflammatory cytokines; p38MAPK, NF-κB/p65, and IDO enzyme expressions; and the concentrations of tryptophan, kynurenine, and serotonin, respectively.ResultOur results showed that POFS was associated with increased expressions of inflammatory cytokines and p38MAPK and higher concentrations of kynurenine and tryptophan on post-operative days 1 and 3; a lower serotonin level on post-operative day 1; and an enhanced translocation of NF-κB/p65 and the IDO enzyme on post-operative days 1, 3, and 5. Ginsenoside Rb1 had an improvement effect on these.ConclusionInflammatory cytokines induced by large abdominal surgery disturb tryptophan metabolism to cause POFS through the activation of the p38MAPK–NF-κB/p65–IDO pathway in the hippocampus. Ginsenoside Rb1 had an anti-fatigue effect on POFS by reducing inflammation and IDO enzyme.

Background: Patients with unresectable locoregional cancer recurrences have limited management options. Reirradiation increases the risk of toxicity, particularly when perilesional dose-volume constraints are exceeded. We present and discuss two cases of previously irradiated tumors in the central nervous system (CNS) that was reirradiated using the pulsed reduced dose-rate radiotherapy (PRDR) technique. Case Description: A 58-year-old female with a history of metastatic small cell lung cancer to the brain status post multiple rounds of radiation and chemotherapy presented with increasing weakness in her right arm and leg. Magnetic resonance imaging (MRI) revealed a growly peripherally enhancing 1.2 cm mass in the left precentral gyrus that had previously received prophylactic cranial irradiation and stereotactic radiosurgery. The patient was re-irradiated with 35 Gy in 100 fractions over 3 weeks, using PRDR with improved motor function at 3-month follow-up. A 41-year-old male with recurrent glioblastoma of the thoracic spinal cord presented with worsening neurological symptoms, including inability to ambulate due to bilateral leg weakness, causing wheelchair use. MRI thoracic spine revealed a recurrent thoracic lesion 2.2 × 1 × 0.8 cm. In addition to chronic chemotherapy, the patient was retreated palliatively in the same area at 50 Gy in 250 fractions, over 6 weeks, using PRDR. The treated lesion was stable on follow-up imaging, and the patient was able to walk with the assistance of a walker. Conclusion: In our two cases, PRDR proved effective in the treatment of recurrent malignant CNS tumors that were previously irradiated. Prospective studies are needed to delineate the efficacy and toxicity of PRDR.

Obstructive sleep apnea (OSA), a common sleep disorder disease, affects millions of people. Without appropriate treatment, this disease can provoke several health-related risks including stroke and sudden death. A variety of treatments have been introduced to relieve OSA. The main present clinical treatments and undertaken research activities to improve the success rate of OSA were covered in this paper. Additionally, guidelines on choosing a suitable treatment based on scientific evidence and objective comparison were provided. This review paper specifically elaborated the clinically offered managements as well as the research activities to better treat OSA. We analyzed the methodology of each diagnostic and treatment method, the success rate, and the economic burden on the world. This review paper provided an evidence-based comparison of each treatment to guide patients and physicians, but there are some limitations that would affect the comparison result. Future research should consider the consistent follow-up period and a sufficient number of samples. With the development of implantable medical devices, hypoglossal nerve stimulation systems will be designed to be smart and miniature and one of the potential upcoming research topics. The transcutaneous electrical stimulation as a non-invasive potential treatment would be further investigated in a clinical setting. Meanwhile, no treatment can cure OSA due to the complicated etiology. To maximize the treatment success of OSA, a multidisciplinary and integrated management would be considered in the future.

Purpose/Objective(s) SIRT2, an NAD+-dependent histone deacetylase, has been shown to play a pivotal role in various physiological processes, however, its role in cancer is currently controversial. In recent years, SIRT2 has been described as both a tumor suppressor and oncogene with divergent expression and function in various malignancies. With established roles in inflammation and glucose and iron metabolism, systemic and microenvironment SIRT2, rather than just intrinsic tumor cell expression, might have implications in tumor progression. Moreover, recent interest in the tumor immune microenvironment has further characterized immune cells involved in tumor suppression, including natural killer (NK) cells. Here, we sought to investigate the biological role of systemic SIRT2 in melanoma tumor progression and to evaluate its interaction with key tumor suppressive immune cells. Materials/Methods Murine models were employed to investigate the effect of systemic SIRT2 on melanoma tumor progression. Tumor progression was measured in terms of weight and volume for 20 days following subcutaneous inoculation of Sirt2-KI and WT mice with B16-F10 melanoma cells. Differences in tumor-infiltrating immune cell composition were determined by flow cytometry. Granzyme B and Ki-67 immunohistochemical staining was used to measure NK cell functional activity and proliferation in in vivo melanoma samples. NK cell function was further evaluated using in vitro and ex vivo mouse models as well as an in vitro human model. Antibody-mediated NK cell depletion established NK cells as an integral player in SIRT2-mediated tumor promotion. Finally, SIRT2 was pharmacologically inhibited by SirReal2, and the effect on melanoma progression and NK cell tumor infiltration was established. Results Increased systemic expression of SIRT2 promoted melanoma progression in mice, resulting in larger and heavier tumors. Mechanistically, systemic SIRT2 overexpression reduced the number of tumor-infiltrating NK cells and suppressed NK cell activation and proliferation within the tumor microenvironment. Furthermore, despite the effect of increased tumor growth rate and tumor volume in wild-type littermate mice, NK cell depletion did not affect that in SIRT2-overexpressing mice. Lastly, pharmacological inhibition of SIRT2 increased NK cell tumor infiltration and suppressed melanoma tumor growth. Conclusion Systemic SIRT2 and NK cell activity exhibit a dynamic functional interaction which promotes melanoma tumor progression. Given the recent renewed interest in NK-cell-mediated immunotherapy response, SIRT2 could represent a new opportunity to mediate immunotherapy response and resistance.

Chemotherapy-induced peripheral neuropathy is among the most common dose-limiting adverse effects of cancer treatment, leading to dose reduction and discontinuation of life-saving chemotherapy and a permanently impaired quality of life for patients. Currently, no effective treatment or prevention is available. Senescence induced during cancer treatment has been shown to promote the adverse effects. Here, we show that cisplatin induces senescent-like neuronal cells in primary culture and in mouse dorsal root ganglia (DRG), as determined by the characteristic senescence markers including senescence-associated beta-galactosidase, accumulation of cytosolic p16INK4A and HMGB1, as well as increased expression of p16Ink4a, p21, and MMP-9. The accumulation of senescent-like neuronal cells in DRG is associated with cisplatin-induced peripheral neuropathy (CIPN) in mice. To determine if depletion of senescent-like neuronal cells may effectively mitigate CIPN, we used a pharmacological ‘senolytic’ agent, ABT263, which inhibits the anti-apoptotic proteins BCL-2 and BCL-xL and selectively kills senescent cells. Our results demonstrated that clearance of DRG senescent neuronal cells reverses CIPN, suggesting that senescent-like neurons play a role in CIPN pathogenesis. This finding was further validated using transgenic p16-3MR mice, which permit ganciclovir (GCV) to selectively kill senescent cells expressing herpes simplex virus 1 thymidine kinase (HSV-TK). We showed that CIPN was alleviated upon GCV administration to p16-3MR mice. Together, the results suggest that clearance of senescent DRG neuronal cells following platinum-based cancer treatment might be an effective therapy for the debilitating side effect of CIPN.

Background Noninvasive monitoring of fetal development and the early detection of pregnancy-associated complications is challenging, largely due to the lack of information about the molecular spectrum during pregnancy. Recently, cell-free DNA in plasma was found to reflect the global nucleosome footprint and status of gene expression, and showed potential for noninvasive health monitoring during pregnancy. Objective(s) We aimed to test the relationships between plasma cell-free DNA profiles and pregnancy biology, and evaluate the use of cell-free DNA profile as a non-invasive method for physiological and pathological status monitoring during pregnancy. Study Design We used genome cell-free DNA sequencing data generated from non-invasive prenatal testing in a total of 2937 pregnant women. For each physiological and pathological conditions, features of cell-free DNA profile were identified using the discovery cohort, and support vector machines classifiers were built and evaluated using independent training and validation cohorts. Results We established nucleosome occupancy profiles at transcription start sites in different gestational trimesters, demonstrated the relationships between gene expression and cell-free DNA coverage at transcription start sites, and showed that the cell-free DNA profiles at transcription start sites represented the biological processes of pregnancy. In addition, using cell-free DNA data, nucleosome profiles of transcription factor binding sites were identified to reflect transcription factor footprint, which may help to reveal the molecular mechanisms underlying pregnancy. Finally, by using machine learning models on low coverage non-invasive prenatal testing data, we evaluated the use of cell-free DNA nucleosome profiles for distinguishing gestational trimesters, fetal gender, and fetal trisomy 21, and highlighted its potential utility for predicting physiological and pathological fetal conditions by using low coverage non-invasive prenatal testing data. Conclusion(s) Our analyses profiled nucleosome footprints and regulatory networks during pregnancy and established a noninvasive, proof-of-principle methodology for health monitoring during pregnancy.

Background: Radio-, chemo- and chemoradio-therapy are essential strategies for cancer therapy. However, there are no reliable non-invasive methods for assessing the effectiveness of these therapies before treatment. Recently, plasma cell-free DNA (cfDNA) was found to reflect the global nucleosome footprint and gene expression status of cancer and immune cells in cancer patients. Therefore, we hypothesized that cfDNA could be used to evaluate the effectiveness of cancer therapy. Methods: In this study, we implemented whole genome sequencing (WGS) of cfDNA collected from patients with three types of cancer before therapy. In addition to promoter profiling, we found that the global chromatin profiles between the sensitive and non-sensitive groups of each cancer type also showed significant changes. To further compare their potentials for effectiveness evaluation of cancer therapy, we developed classifiers based on the WGS data derived from 194 locally advanced rectal cancer (LARC) patients with multiple machine learning models for identifying LARC with pathological complete response (pCR). Findings: We found that not only the promoter profiling but also some variables of global chromatin profiles of cfDNA between the sensitive and non-sensitive groups of cancer patients showed significant difference. By comparing their performance for effectiveness evaluation, we found that the predictive classifier based on promoter profiling with the SVM model (PPCET) had the largest AUC of 0.89 (95% CI 0.83-0.94), which may highlight the potential applications of promoter profiling as a novel non-invasive in vivo approach for assessing the effectiveness of cancer therapy. Interpretation: PPCET is a non-invasive technique for predicting the effectiveness of cancer therapy before treatment that may help to prevent indiscriminate use of drugs, reduce toxicity and side effects, and improve the curative effect and quality of life. Funding Statement: This work was supported by National Natural Science Foundation of China [81802435]; China Postdoctoral Science Foundation funded project [2019T120742, 2016M602486]; Natural Science Foundation of Guangdong Province [2018A030313286, 2015B020233009]; Science and Technology Program of Guangzhou [201604020104, 201803040009]. Declaration of Interests: The authors declare that they have no conflicts of interest. Ethics Approval Statement: All plasma samples were obtained under Institutional Review Board approved protocols with informed consent from all participants for research use.

PARP inhibitors (PARPi) are potentially effective therapeutic agents capable of inducing synthetic lethality in tumors with deficiencies in homologous recombination (HR)-mediated DNA repair such as those carrying BRCA1 mutations. However, BRCA mutations are rare, the majority of tumors are proficient in HR repair, and thus most tumors are resistant to PARPi. Previously, we observed that ionizing radiation (IR) initiates cytoplasmic translocation of BRCA1 leading to suppression of HR-mediated DNA repair and induction of synthetic PARPi lethality in wild-type BRCA1 and HR-proficient tumor cells. The tumor suppressor p53 was identified as a key factor that regulates DNA damage–induced BRCA1 cytoplasmic sequestration following IR. However, the role of p53 in IR-induced PARPi sensitization remains unclear. This study elucidates the role of p53 in IR-induced PARPi cytotoxicity in HR-proficient cancer cells and suggests p53 status may help define a patient population that might benefit from this treatment strategy. Sensitization to PARPi following IR was determined in vitro and in vivo utilizing human breast and glioma tumor cells carrying wild-type BRCA1 and p53, and in associated cells in which p53 function was modified by knockdown or mutation. In breast and glioma cells with proficient HR repair, IR-induced BRCA1 cytoplasmic sequestration, HR repair inhibition, and subsequent PARPi sensitization in vitro and in vivo was dependent upon functional p53. Implications: Implications: p53 status determines PARP inhibitor sensitization by ionizing radiation in multiple BRCA1 and HR-proficient tumor types and may predict which patients are most likely to benefit from combination therapy. Mol Cancer Res; 16(7); 1092–102. ©2018 AACR.

Robust optimization generates scenario-based plans by a minimax optimization method to find optimal scenario for the trade-off between target coverage robustness and organ-at-risk (OAR) sparing. In this study, 20 lung cancer patients with tumors located at various anatomical regions within the lungs were selected and robust optimization photon treatment plans including intensity modulated radiotherapy (IMRT) and volumetric modulated arc therapy (VMAT) plans were generated. The plan robustness was analyzed using perturbed doses with setup error boundary of ±3 mm in anterior/posterior (AP), ±3 mm in left/right (LR), and ±5 mm in inferior/superior (IS) directions from isocenter. Perturbed doses for D99 , D98 , and D95 were computed from six shifted isocenter plans to evaluate plan robustness. Dosimetric study was performed to compare the internal target volume-based robust optimization plans (ITV-IMRT and ITV-VMAT) and conventional PTV margin-based plans (PTV-IMRT and PTV-VMAT). The dosimetric comparison parameters were: ITV target mean dose (Dmean ), R95 (D95 /Dprescription ), Paddick's conformity index (CI), homogeneity index (HI), monitor unit (MU), and OAR doses including lung (Dmean , V20 Gy and V15 Gy ), chest wall, heart, esophagus, and maximum cord doses. A comparison of optimization results showed the robust optimization plan had better ITV dose coverage, better CI, worse HI, and lower OAR doses than conventional PTV margin-based plans. Plan robustness evaluation showed that the perturbed doses of D99 , D98 , and D95 were all satisfied at least 99% of the ITV to received 95% of prescription doses. It was also observed that PTV margin-based plans had higher MU than robust optimization plans. The results also showed robust optimization can generate plans that offer increased OAR sparing, especially for normal lungs and OARs near or abutting the target. Weak correlation was found between normal lung dose and target size, and no other correlation was observed in this study.

Platinum-based chemotherapies, such as cisplatin, play a large role in cancer treatment. The development of resistance and treatment toxicity creates substantial barriers to disease control, yet. To enhance the therapeutic index of cisplatin-based chemotherapy, it is imperative to circumvent resistance and toxicity while optimizing tumor sensitization. One of the primary mechanisms by which cancer cells develop resistance to cisplatin is through upregulation of DNA repair pathways. In this review, we discuss the DNA damage response in the context of cisplatin-induced DNA damage. We describe the proteins involved in the pathways and their roles in resistance development. Common biomarkers for cisplatin resistance and their utilization to improve patient risk stratification and treatment personalization are addressed. Finally, we discuss some of the current treatments and future strategies to circumvent the development of cisplatin resistance.

We hypothesized that sorafenib (BAY 43-9006), an oral multi-kinase inhibitor, used in combination with SRS will improve overall intracranial control. This Phase I study assesses the safety, tolerability, and maximal tolerated dose of sorafenib administered with SRS to treat 1-4 brain metastases. This was an open label phase I dose escalation study with an expansion cohort. Eligible adults had 1-4 brain metastases from solid malignancies. Sorafenib was begun 5-7 days prior to SRS and continued for 14 days thereafter. Dose escalation of sorafenib was conducted via a "3 + 3" dose escalation design. Dose limiting toxicities (DLT) were determined 1 month after SRS and defined as ≥grade 3 neurologic toxicities. Twenty-three patients were enrolled. There were no DLTs at dose level 1 (400 mg per day) or dose level 2 (400 mg twice per day). An expansion cohort of 17 patients was treated at dose level 2. There were six grade 3 toxicities: hypertension (n = 2), rash (n = 1), lymphopenia (n = 1), hypokalemia (n = 1), fatigue (n = 1) and hand-foot syndrome (n = 1). All of these were attributable to sorafenib and not to the combination with SRS. The median time to CNS progression was 10 months, 1 year CNS progression-free survival was 46%, the median overall survival was 11.6 months and the 1 year overall survival was 46%. The use of sorafenib concurrent with SRS for the treatment of 1-4 brain metastases is safe and well tolerated at 400 mg twice a day. Our recommended phase II dose of concurrent sorafenib with SRS would be 400 mg twice daily.

Automatic machine learning (automl) is a crucial technology in machine learning. Sequential model-based optimisation algorithms (SMBO) (e.g., SMAC, TPE) are state-of-the-art hyperparameter optimisation methods in automl. However, SMBO does not consider known information, like the best hyperparameters high possibility range and gradients. In this paper, we accelerate the traditional SMBO method and name our method as accSMBO. In accSMBO, we build a gradient-based multikernel Gaussian process with a good generalisation ability and we design meta-acquisition function which encourages that SMBO puts more attention on the best hyperparameters high possibility range. In L2 norm regularised logistic loss function experiments, our method exhibited state-of-the-art performance.

Effect of laparoscopic myomectomy on the serum levels of interleukin-6 (IL-6) and total antioxidant capacity (TAC), and the ovarian function of patients with uterine fibroids was studied. Ninety patients with uterine fibroids admitted to The Central Hospital of Wuhan, Tongji Medical College, Huazhong University of Science and Technology, from March 2013 to July 2014, were randomly divided into the experimental group (n=45) and the control group (n=45). The experimental group was treated with laparoscopic myomectomy, and the control group was treated with abdominal myomectomy. ELISA was used for detecting IL-6 and TAC content. Radioimmunoassay (RIA) was used for detecting serum ovarian function indicators, including estradiol (E2), follicle-stimulating hormone (FSH) and luteinizing hormone (LH). Patients in the experimental group had significantly lower intraoperative blood loss, postoperative exhaust time, number of days to recovery and return to work, number of intraoperative fibroids removed, and length of stay than those in the control group (P

PARP inhibitors are one of the success stories of targeted cancer therapy. In the last few years, these drugs have been approved by the US Food and Drug Administration (FDA) for the treatment of breast and ovarian cancers. PARP inhibitors are useful in the treatment of DNA double-strand break repair deficient tumors such as those with BRCA1 or BRCA2 mutations. In this chapter, we discuss the pathophysiology of breast and ovarian cancers in association with DNA repair and genomic instability. We focus our discussion on the use of PARP inhibitors in these malignancies. We also discuss how the tumors gain resistance to these agents, including utilizing strategies such as restoration of homologous recombination-mediated DNA double-strand break repair pathway and stabilization of replication forks. We review possible approaches for overcoming resistance to PARP inhibitors including targeting protein kinases and alternate signaling pathways, exploiting cell cycle regulation, and drug pumps. We end with the benefits of novel therapies, their limitations and work that remains to be done.

SIRT2, an NAD+-dependent histone deacetylase, has been shown to play a pivotal role in various physiological processes, however, its role in cancer is currently controversial. In recent years, SIRT2 has been described as both a tumor suppressor and oncogene with divergent expression and function in various malignancies. Using murine xenograft melanoma models, our results suggest increased systemic expression of SIRT2 promotes tumor progression. In this study, SIRT2-overexpressing mice exhibited enhanced tumor growth and larger tumor volumes compared to their wild-type littermates. Mechanistically, systemic overexpression of SIRT2 reduces the number of tumor-infiltrating natural killer (NK) cells and suppresses NK cell activation and proliferation within tumor microenvironment (TME). Furthermore, despite the effect of increased tumor growth rate and tumor volume in wild-type littermate mice, NK cell depletion does not affect that in SIRT2-overexpressing mice. Lastly, pharmacological inhibition of SIRT2 increases NK cell tumor infiltration and suppresses melanoma tumor growth in mice. The findings of this study identify a dynamic functional interaction between systemic SIRT2 and NK cell activity, which promotes melanoma tumor progression. Given the recent renewed interest in NK-cell-mediated immunotherapy response, SIRT2 could present a new opportunity to mediate immunotherapy response and resistance. Citation Format: Fen Xia. Sirt2: A tumor suppressor or promotor? [abstract]. In: Proceedings of the AACR Virtual Special Conference on Radiation Science and Medicine; 2021 Mar 2-3. Philadelphia (PA): AACR; Clin Cancer Res 2021;27(8_Suppl):Abstract nr IA-024.

Platinum-based chemotherapy-induced peripheral neuropathy (CIPN) is one of the most common dose-limiting effects of cancer treatment and results in dose reduction and discontinuation of life-saving chemotherapy. Its debilitating effects are often permanent and lead to lifelong impairment of quality of life in cancer patients. While the mechanisms underlying the toxicity are not yet fully defined, dorsal root ganglia sensory neurons play an integral role in symptom development. DNA-platinum adducts accumulate in these cells and inhibit normal cellular function. Nucleotide excision repair (NER) is integral to the repair of platinum adducts, and proteins involved in its mechanism serve as potential targets for future therapeutics. This review aims to highlight NER’s role in cisplatin-induced peripheral neuropathy, summarize current clinical approaches to the toxicity, and discuss future perspectives for the prevention and treatment of CIPN.

To investigate whether resistin is associated with early atherosclerosis in male smokers. The present study consecutively enrolled 50 male smokers. Their serum resistin contents were detected with enzyme linked immunosorbent assay (ELISA), and subclinical atherosclerosis indices, including carotid inner middle thickness (IMT) and arterial elasticity indices (C1 and C2), were measured. The association between serum resistin levels and IMT, C1 and C2 were respectively evaluated with the Pearson's correlation coefficient method. The results showed that the serum resistin level had a positive association with IMT (r = 0.307, p = .030), but were both inversely associated with C1 (r = -0.440, p = .001) and C2 (r = -0.381, p = .006). These associations remained significant even after adjustment for cardiovascular confounders. In conclusion, serum resistin concentration was independently associated with early atherosclerosis in male smokers.

Esophageal squamous cell carcinoma (ESCC) is one of the most common cancers in China with very low 5-year survival rate mostly due to the paucity of effective early diagnostic methods. Serum autoantibodies against 9 tumor-associated antigens (TAAs) from ESCC patients and healthy controls were detected by enzyme-linked immunosorbent assay to evaluate their performances in the immunodiagnosis of ESCC. Logistic regression models were generated to predict the probability of individuals being diagnosed with ESCC in training cohort (648 participants) and further validated in another independent cohort (372 participants). Finally, a panel of four TAAs showed high diagnostic accuracy with areas under the receiver operating characteristic curve of 0.838 in training cohort and 0.872 in validation cohort, respectively. The percentages of individuals correctly classified were 77.01 % in training cohort and 78.49 % in validation cohort, respectively. This model could discriminate early-stage (AJCC stage 0, I and II) ESCC patients from normal controls, with true-positive rate (TPR) of 67.57 % in training cohort and TPR of 63.33 % in validation cohort, and the overall TPR for early-stage ESCC was 66.85 % when the two cohorts were combined. The diagnostic performance of this model showed no significant difference between early-stage and late-stage (AJCC stage III and IV) ESCC patients. In summary, the optimized model with 4 TAAs has a high diagnostic performance for ESCC detection, especially for early-stage ESCC.

Learning to rank has received great attention in recent years as it plays a crucial role in many applications such as information retrieval and data mining. The existing concept of learning to rank assumes that each training instance is associated with a reliable label. However, in practice, this assumption does not necessarily hold true as it may be infeasible or remarkably expensive to obtain reliable labels for many learning to rank applications. Therefore, a feasible approach is to collect labels from crowds and then learn a ranking function from crowdsourcing labels. This study explores the listwise learning to rank with crowdsourcing labels obtained from multiple annotators, who may be unreliable. A new probabilistic ranking model is first proposed by combining two existing models. Subsequently, a ranking function is trained by proposing a maximum likelihood learning approach, which estimates ground-truth labels and annotator expertise, and trains the ranking function iteratively. In practical crowdsourcing machine learning, valuable side information (e.g., professional grades) about involved annotators is normally attainable. Therefore, this study also investigates learning to rank from crowd labels when side information on the expertise of involved annotators is available. In particular, three basic types of side information are investigated, and corresponding learning algorithms are consequently introduced. Further, the top-k learning to rank from crowdsourcing labels are explored to deal with long training ranking lists. The proposed algorithms are tested on both synthetic and real-world data. Results reveal that the maximum likelihood estimation approach significantly outperforms the average approach and existing crowdsourcing regression methods. The performances of the proposed algorithms are comparable to those of the learning model in consideration reliable labels. The results of the investigation further indicate that side information is helpful in inferring both ranking functions and expertise degrees of annotators.

Listwise learning to rank (LTR) is aimed at constructing a ranking model from listwise training data to order objects. In most existing studies, each training instance consists of a set of objects described by preference features. In a preference feature space for the objects in training, the structure of the objects is associated with the absolute preference degrees for the objects. The degrees significantly influence the ordering of the objects. Nevertheless, the structure of the training objects in their preference feature space has rarely been studied. In addition, most listwise LTR algorithms yield a single linear ranking model for all objects, but this ranking model may be insufficient to capture the underlying nonlinear ranking mechanism among all objects. This study proposes a divide-and-train method to learn a nonlinear ranking model from listwise training data. First, a rank-preserving clustering approach is used to infer the structure of objects in their preference feature space and all the objects in training data are divided into several clusters. Each cluster is assumed to correspond to a preference degree and an ordinal regression function is then learned. Second, considering that relations exist among the clusters, a multi-task listwise ranking approach is then employed to train linear ranking functions for all the clusters (or preference degrees) simultaneously. Our proposed method utilizes both the (relative) preferences among objects and the intrinsic structure of objects. Experimental results on benchmark data sets suggest that the proposed method outperforms state-of-the-art listwise LTR algorithms.

Introduction The popularity of online physician rating websites has risen substantially. These third-party sites have the potential to significantly influence patients’ perception of their healthcare providers. The purpose of this study was to evaluate online ratings of U.S. radiation oncologists (ROs) on Vitals.com, one of the most popular physician rating websites, and the variables that most significantly affect patients’ overall rating (OR) of their ROs. Methods The Centers for Medicare and Medicaid Physician Comparable Downloadable File was analyzed to obtain data on all self-identified ROs in the U.S. and Puerto Rico. Patient Review Satisfaction Scores (PRSS) that ranged from one (poor) to five (excellent) for the following variables were recorded: OR, accurate diagnosis, spending appropriate time with patients, ease of appointment, courteous staff, bedside manner, follow-up after visit, promptness, and wait time. Associations among these factors were assessed. Results Of 4,443 self-identifying Medicare-accepting ROs, 1,797 (40.4%) ROs who had at least one OR rating and at least one written comment were included in this study. The ROs’ mean OR was 4.34 ± 0.2 (median 4; 30% received a score of 5; 78% received a score greater than 4). OR was found to have a strong correlation with accuracy of diagnosis (r = 0.69), bedside manner (r = 0.71), and spends appropriate time with patients (r = 0.69). With the exception of the number of ratings (p = 0.07), physicians with over 10 years of experience showed statistically significant differences in how much better they scored in each of the variables compared to those with less than 10 years of experience (p < 0.01 for each characteristic). Significant differences in OR were also observed between ROs whose wait times exceeded 20 minutes compared to those with wait times less than 10 minutes (p < 0.01) for all internal and external metrics except for the number of ratings (p = 0.42) and number of reviews (p = 0.88) Conclusion Patients are providing high ratings for their ROs on Vitals.com and are more frequently recommending them to friends and family. Given the rise in popularity of third-party physician rating sites, it is important for ROs to understand the various factors that may influence their online ratings.

In this letter, we propose a dual-band filter using open stub-loaded resonators. The dual-band filter is analysed by transmission line theory and the physical dimensions of the structure can be easily deduced for 5G application. Thus, based on full-wave simulation, the proposed design is verified.

In this paper, a dual-band notched compact UWB antenna fed by CPW is presented. It is printed on FR4 substrate with a compact size of 20mm*29.5mm*1.5mm. It is designed to cover the band from 2.883 GHz to 20 GHz. This antenna is able to reduce the interference at IEEE 802.16 WiMAX (3.3-3.8GHz) band and IEEE 802.11 WLAN (5.15-5.825GHz) band, whose two frequency notches are realized by embedding split ring resonator (SRR) in the coplanar waveguides and an inverted U-shaped slot on the antenna. It has an omnidirectional radiation pattern and the gain is suppressed well at the two notched bands. This proposed work has wider bandwidth and good band rejection performance in comparison with the state of arts.

Platinum-based chemotherapy–induced peripheral neuropathy is one of the most common causes of dose reduction and discontinuation of life-saving chemotherapy in cancer treatment; it often causes permanent impairment of quality of life in cancer patients. The mechanisms that underlie this neuropathy are not defined, and effective treatment and prevention measures are not available. Here, we demonstrate that SIRT2 protected mice against cisplatin-induced peripheral neuropathy (CIPN). SIRT2 accumulated in the nuclei of dorsal root ganglion sensory neurons and prevented neuronal cell death following cisplatin treatment. Mechanistically, SIRT2, an NAD(+)-dependent deacetylase, protected neurons from cisplatin cytotoxicity by promoting transcription-coupled nucleotide excision repair (TC-NER) of cisplatin-induced DNA cross-links. Consistent with this mechanism, pharmacological inhibition of NER using spironolactone abolished SIRT2-mediated TC-NER activity in differentiated neuronal cells and protection of neurons from cisplatin-induced cytotoxicity and CIPN in mice. Importantly, SIRT2’s protective effects were not evident in lung cancer cells in vitro or in tumors in vivo. Taken together, our results identified SIRT2’s function in the NER pathway as a key underlying mechanism of preventing CIPN, warranting future investigation of SIRT2 activation–mediated neuroprotection during platinum-based cancer treatment.

Methionine dependency describes the characteristic rapid in vitro death of most tumor cells in the absence of methionine. Combining chemotherapy with dietary methionine deprivation [methionine-deficient diet (MDD)] at tolerable levels has vast potential in tumor treatment; however, it is limited by MDD-induced toxicity during extended deprivation. Recent advances in imaging and irradiation delivery have created the field of stereotactic body radiotherapy (SBRT), where fewer large-dose fractions delivered in less time result in increased local-tumor control, which could be maximally synergistic with an MDD short course. Identification of the lowest effective methionine dietary intake not associated with toxicity will further enhance the cancer therapy potential. In this study, we investigated the effects of MDD and methionine-restricted diet (MRD) in primary and metastatic melanoma models in combination with radiotherapy (RT). In vitro, MDD dose-dependently sensitized mouse and human melanoma cell lines to RT. In vivo in mice, MDD substantially potentiated the effects of RT by a significant delay in tumor growth, in comparison with administering MDD or RT alone. The antitumor effects of an MDD/RT approach were due to effects on one-carbon metabolism, resulting in impaired methionine biotransformation via downregulation of Mat2a, which encodes methionine adenosyltransferase 2A. Furthermore, and probably most importantly, MDD and MRD substantially diminished metastatic potential; the antitumor MRD effects were not associated with toxicity to normal tissue. Our findings suggest that modulation of methionine intake holds substantial promise for use with short-course SBRT for cancer treatment.

Previous studies normally formulate Chinese word segmentation as a character sequence labeling task and optimize the solution in sentence-level. In this paper, we address Chinese word segmentation as a document-level optimization problem. First, we apply a state-of-the-art approach, i.e., long short-term memory (LSTM), to perform character classification; Then, we propose a global objective function on the basis of character classification and achieve global optimization via Integer Linear Programming (ILP). Specifically, we propose several kinds of global constrains in ILP to capture various segmentation knowledge, such as segmentation consistency and domain-specific regulations, to achieve document-level optimization, besides label transition knowledge to achieve sentence-level optimization. Empirical studies demonstrate the effectiveness of the proposed approach to domain-specific Chinese word segmentation.

As the main revenue source of Internet companies, online advertising is always a significant topic, where click-through rate (CTR) prediction plays a central role. In online advertising systems, there are often many advertisement products. Due to the competition in the bidding mechanism, some advertising products may get lots of data to train the CTR prediction model while some may lack high-quality data. However, to predict accurate CTR, a large amount of data is needed. Therefore, transfer knowledge from the large product (source) to the small product (target) is necessary. We propose a transfer learning method that iteratively updates the data weights to selectively combine source data with target data for training. To efficiently process huge advertisement data, we design a sampling strategy based on the gradient information, and implement the algorithm with a MapReduce-like machine learning framework. We do experiments on real advertisement datasets. The results show that our approach improves the accuracy of CTR prediction compared to the supervised learning method.

Introduction Although tumor metastases remain significant drivers of mortality, the genetic factors that increase the risks of metastases are not fully identified. Interleukin 6 (IL-6) has emerged as an important factor in breast cancer progression with IL-6 single nucleotide polymorphism (SNP) variants shown to affect survival. We hypothesized that SNPs of the IL-6 promoter at rs1800795 in breast cancer patients are associated with distant metastases. Methods We performed an initial case-control study using Vanderbilt University Medical Center's BioVU, a genomic biobank linked to de-identified electronic medical records in the Synthetic Derivative database, to identify germline SNPs that may predict the development of metastatic disease to any site from any solid tumor including breast cancer. We identified a SNP in IL-6: rs1800795 to be of significance and evaluated this finding using a separate, matched-pair cohort of breast cancer patients with and without metastases from The Ohio State University Wexner Medical Center. Results The initial study suggested that GG relative to CG at rs1800795 (OR 1.52; 95% CI 1.14-2.02; p = 0.004) was significantly associated with the development of metastases. This association was also observed in the Ohio State University cohort (OR 2.23; 95% CI 1.06-4.71; p = 0.001). There were no significant relationships between rs1800795 status and any patient or tumor characteristics, including estrogen receptor status. Conclusions These findings suggest that GG SNP at IL-6: rs1800795 may indicate an increased risk of metastasis of primary breast cancer. Further studies in larger population sets are warranted as advanced screening and prophylactic intervention might be employed in GG carriers.

The difficulty of getting medical treatment is one of major livelihood issues in China. Since patients lack prior knowledge about the spatial distribution and the capacity of hospitals, some hospitals have abnormally high or sporadic population densities. This paper presents a new model for estimating the spatiotemporal population density in each hospital based on location-based service (LBS) big data, which would be beneficial to guiding and dispersing outpatients. To improve the estimation accuracy, several approaches are proposed to denoise the LBS data and classify people by detecting their various behaviors. In addition, a long short-term memory (LSTM) based deep learning is presented to predict the trend of population density. By using Baidu large-scale LBS logs database, we apply the proposed model to 113 hospitals in Beijing, P. R. China, and constructed an online hospital recommendation system which can provide users with a hospital rank list basing the real-time population density information and the hospitals' basic information such as hospitals' levels and their distances. We also mine several interesting patterns from these LBS logs by using our proposed system.

Optimal treatment of metastases to the central nervous system (CNS) in patients with malignant melanoma remains a clinical challenge. In particular, for patients with BRAF-mutant melanoma and CNS metastases, much remains unknown about the safety and efficacy of the novel BRAF-targeted agents when administered in close sequence with radiation. We report two cases of rapid development of CNS radiation necrosis in patients with metastatic melanoma treated with the BRAF inhibitor, vemurafenib, closely sequenced with stereotactic radiosurgery or fractionated stereotactic radiation therapy. In the absence of prospective safety data from clinical trials, we advise vigilance in monitoring patients with BRAF-mutant melanoma whose treatment plan includes CNS radiation and vemurafenib and caution when assessing treatment response within the CNS in these patients.

In this study, we present an organic field-effect transistor floating-gate memory using polysilicon (poly-Si) as a charge trapping layer. The memory device is fabricated on a N+–Si/SiO2 substrate. Poly-Si, polymethylmethacrylate, and pentacene are used as a floating-gate layer, tunneling layer, and active layer, respectively. The device shows bidirectional storage characteristics under the action of programming/erasing (P/E) operation due to the supplied electrons and holes in the channel and the bidirectional charge trapping characteristic of the poly-Si floating-gate. The carrier mobility and switching current ratio (I on/I off ratio) of the device with a tunneling layer thickness of 85 nm are 0.01 cm 2 · V − 1 · s − 1 and 102, respectively. A large memory window of 9.28 V can be obtained under a P/E voltage of ±60 V.

Sparse representation is a new image representation theory. It can accurately represent the image information. In this paper, a novel fusion scheme using sparse representation is proposed. The sparse representation is conducted on overlapping patches. Each source image is divided into patches, and all the patches are transformed into vectors. Decompose the vectors into theirs sparse representations using orthogonal matching pursuit. Sparse coefficients are fused with the maximum absolute. The simulation results show that the proposed method can provide high-quality images.

An image fusion algorithm by combing Nonsubsampled Contourlet Transform and Pulse Coupled Neural Network is proposed in this paper. Two primitive matched images are decomposed in NSCT domain. The low frequency used the regional energy weighted fusion. The high frequency coefficients are input to the PCNN. Finally, the fused result is obtained through inverse NSCT. The novel algorithm will employ features and contain the most of the energy of image. The simulation results show that the results are much better than the experimental results, which the mutual information and QAB/F parameters are higher than the contrast method of fusion results.

In recent years, there have been multiple breakthroughs in our understanding of lung cancer biology. Despite significant advances in molecular targeted therapies, DNA-damaging cytotoxic therapies will remain the mainstay of lung cancer management for the near future. Similar to the concept of personalized targeted therapies, there is mounting evidence that perturbations in DNA repair pathways are common in lung cancers, altering the resistance of the affected tumors to many chemotherapeutics as well as radiation. Defects in DNA repair may be due to a multitude of mechanisms including gene mutations, epigenetic events, and alterations in signal transduction pathways such as epidermal growth factor receptor and phosphoinositide 3-kinase/AKT. Functional biomarkers that assess the subcellular localization of central repair proteins in response to DNA damage may prove useful for individualization of cytotoxic therapies including poly(adenosine diphosphate-ribose) polymerase inhibitors. A better mechanistic understanding of cellular sensitivity and resistance to DNA damaging agents should facilitate the development of novel, individualized treatment approaches. Absolute resistance to radiation therapy, however, does not exist. To some extent, radiation therapy will always have to remain unselective and indiscriminant to eradicate persistent, drug-resistant tumor stem cell pools.

This paper briefly reviews our recent analytical and experimental results on 3 interrelated features beyond the peak load in heterogeneous media: continuous bifurcation, damage localization and catastrophic rupture (CR). Firstly, an Elastic Statistically-Brittle model (ESB) was introduced to formulate the basic features of a kind of heterogeneous media, like rocks and cements. The global mean field approximation (GMF) shows that the measure of heterogeneity, like the Weibull modulus m in the distribution of meso-strength plays a key role to distinguish CR from gradual failure. Then, with the ESB model and corresponding experimental results, continuous bifurcation and damage localization are discussed. In accord with these, regional mean field approximation (RMF) is adopted and it shows that any scale of damage localization can satisfy the conservation laws in continuum mechanics. This implies that catastrophic rupture could appear at any state beyond the peak load, depending on the unknown evolution of damage localization zone. Hence, catastrophic rupture seems to occur stochastically at macroscopic level. On the other hand, both experimental and analytic studies demonstrate that a robust power law singularity (-1/2) appears ahead of CR. Preliminary applications of these ideas are briefly described.

A bandpass differentiator based on network synthesis methods is designed in this paper. The synthesis methods can implement arbitrary response with frequency. If the transfer function S21 is specified, the stability of the differentiator can be ensured absolutely by chosen the left half-plane roots of F(s). The K/J inverter is designed to implement the coupled circuit. Then, a microwave differentiator is designed and simulated. The simulated results of the differentiator show good agreement with the theoretical analysis.

Estrogen receptor-a (ER) protein plays a key role in breast carcinogenesis, and common genetic variants in the corresponding gene locus have been associated with breast cancer risk in different populations. Here, we analyzed estrogen receptor 1 (ESR1) associations in two hospital-based studies of patients from the south of China. Three single-nucleotide polymorphisms (SNPs; rs3757318, rs2046210, and rs3734805) in ESR1 were selected from previous genome-wide association study results and were genotyped using the Sequenom MassARRAY® iPLEX System in 845 breast cancer patients and 882 healthy controls. Association analysis based on unconditional logistic regression was carried out to determine the odds ratio (OR) and 95% confidence interval (95%CI) for each SNP. Stratified analyses according to the status of ER and progesterone receptor (PR) were also performed. Of the three SNPs, rs3757318 did not pass the Hardy-Weinberg equilibrium test and was excluded from the subsequent analysis. The other two SNPs (rs2046210 and rs3734805) were strongly associated with susceptibility to breast cancer. Allele T of rs2046210 and allele C of rs3734805 were risk alleles and the adjusted ORs were 1.348 (95%CI = 1.172-1.550, P = 0.0001) and 1.319 (95%CI = 1.144-1.522, P = 0.0001), respectively. Furthermore, the risk allele of rs2046210 gave negative results for ER and PR expression in an immunohistochemical test, with ORs of 0.602 (95%CI = 0.384-0.944, P = 0.027) and 0.532 (95%CI = 0.338-0.837, P = 0.006), respectively. Our study further supports associations between rs2046210 and rs3734805 and breast cancer risk in Chinese women.

Genome-wide association studies (GWAS) have identified 11 loci that influence the risk of developing colorectal cancer (CRC). Given that these studies were conducted in European Caucasian populations, it is not clear whether the results are relevant for populations with different ethnicities. The aim of this study was to examine these associations in a southern Chinese population.Eleven single-nucleotide polymorphisms (SNPs), rs12701937, rs16892766, rs7014346, rs6983267, rs719725, rs10795668, rs3802842, rs4444235, rs9929218, rs10411210, and rs961253, were genotyped in 229 CRC patients and 267 controls using the MassArray SNP genotyping system.Evidence of an association with CRC was found for four of the 11 loci. The strongest associations were with rs4444235 and rs961253, with significant odds ratios close to those reported in previous GWAS. Among these four loci, rs719725 and rs4444235 were significantly associated with female gender, rs3802842, rs961253, and rs4444235 with early disease onset, and rs3802842 with later disease onset. However, no associations with CRC risk were detected for six other loci (rs9929218, rs10411210, rs12701937, rs7014346, rs6983267, and rs10795668), and one SNP, rs16892766, was not polymorphic in any of the study participants.The rs4444235 and rs961253 loci are strongly associated with the risk of CRC in southern Chinese.

A genome-wide study has shown an association between SNPs located on 17q21 and asthma. Such associations have been identified in several populations, but little is known about the Han Chinese population. We conducted a case-control study in a Han Chinese population to investigate the relationship between SNPs located on 17q21 and asthma; 241 asthmatic patients and 212 healthy controls were recruited from the outpatient clinics of the Nanfang Hospital, Guangdong Province, southern China. We genotyped six SNPs (rs8067378, rs8069176, rs2305480, rs4795400, rs12603332, and rs11650680) located on 17q21 with the Sequenom MassARRAY iPLEX platform. For two of these six loci (rs2305480 and rs8067378), there was evidence of association with asthma, and there was a weak association of asthma with rs8069176. We confirm that genetic variants on 17q21 are associated with asthma in the Han Chinese population.

The Wnt signaling pathway is crucial for pulmonary development and differentiation; dysregulation of the Wnt signaling pathway may impair lung function. Indeed, single nucleotide polymorphisms (SNPs) of Wnt pathway-related genes have been suggested as risk factors for certain types of cancers. In this study, we aimed to evaluate the influence of SNPs in Wnt-related genes (TCF2, MMP9) on susceptibility to lung cancer.Polymorphisms of TCF2 rs4430796, MMP9 rs2250889, and MMP9 rs17576 were studied in Han Chinese subjects, including 135 patients with lung cancer and 176 controls, using the Sequenom MassARRAY platform. The association of genotypes with susceptibility to lung cancer was analyzed using odds ratio (OR), with 95% confidence interval (95% CI) and χ(2).The three SNPs (rs4430796, rs2250889, and rs17576) were found to be significantly associated with an increased risk of lung cancer. The AA genotype and AG+AA genotype of rs4430796 showed a significantly increased susceptibility to lung cancer compared with the GG genotype (adjusted OR=6.03, 95% CI: 1.30-28.09, P=0.022; 5.55, 95% CI: 1.20-25.58, P=0.028). Compared with the rs17576 GG genotype, the AG and AG+AA genotypes were also associated with a significant risk (adjusted OR=2.65, 95% CI: 1.60-4.37, P≤0.001; 2.57, 95% CI: 1.59-4.19, P≤0.001) whereas the rs2250889 CG and CG+GG genotypes had 2.97-fold (95% CI: 1.81-4.85; P≤0.001) and 2.80-fold increased associations with lung cancer (95% CI: 1.73-4.54; P≤0.001), respectively, compared with the rs2250889 CC genotype. Furthermore, the association of rs4430796 with lung cancer became insignificant (P0.05) after adjusting for gender and rs2250889.The three SNPs may play a role in the predisposition of members of the Han Chinese population to lung cancer.

Collaborative social annotation systems allow users to record and share their original keywords or tag attachments to Web resources such as Web pages, photos, or videos. These annotations are a method for organizing and labeling information. They have the potential to help users navigate the Web and locate the needed resources. However, since annotations are posted by users under no central control, there exist problems such as spam and synonymous annotations. To efficiently use annotation information to facilitate knowledge discovery from the Web, it is advantageous if we organize social annotations from semantic perspective and embed them into algorithms for knowledge discovery. This inspires the Web page recommendation with annotations, in which users and Web pages are clustered so that semantically similar items can be related. In this paper we propose four graphic models which cluster users, Web pages and annotations and recommend Web pages for given users by assigning items to the right cluster first. The algorithms are then compared to the classical collaborative filtering recommendation method on a real-world data set. Our result indicates that the graphic models provide better recommendation performance and are robust to fit for the real applications.

Ranking problems have recently become an important research topic in the joint field of machine learning and information retrieval. This paper presented a new splitting rule that introduces a metric, i.e., an impurity measure, to construct decision trees for ranking tasks. We provided a theoretical basis and some intuitive explanations for the splitting rule. Our approach is also meaningful to collaborative filtering in the sense of dealing with categorical data and selecting relative features. Some experiments were made to illustrate our ranking approach, whose results showed that our algorithm outperforms both perceptron-based ranking and the classification tree algorithms in term of accuracy as well as speed.

PURPOSE: Pyruvate kinase muscle isoenzyme 2 (PKM2) is a key enzyme in aerobic glycolysis and is thought to contribute to cancer cell metabolic reprogramming. The aim of this study was to evaluate PKM2 immunohistochemical expression as a potential prognostic biomarker in pancreatic ductal adenocarcinoma (PDAC). METHODS: A tissue microarray was constructed using surgical specimens for 115 patients who underwent resections for PDAC, stained with PKM2 antibody, and scored for expression level. Statistical analyses were performed to investigate the association between PKM2 and patient survival, tumor stage, tumor grade, surgical margin status, lymph node ratio, perineural invasion status, or the use of adjuvant chemotherapy. RESULTS: Fifty-three percent of tumors had positive PKM2 expression, and 47 % of tumors had negative PKM2 expression. PKM2 expression was associated with overall survival (HR 0.56, p=0.007) and CA 19–9 levels (p=0.035), but was not associated with tumor stage, tumor grade, surgical margin status, lymph node ratio, perineural invasion, or adjuvant chemotherapy use. CONCLUSIONS: PKM2 expression is associated with overall survival in PDAC. Further studies are warranted to validate the value of PKM2 as a prognostic biomarker and to examine the potential utility of PKM2 in predicting treatment response, as well as a potential therapeutic target in PDAC.