Pipelined biomedical event extraction rivaling joint learning.

Biomedical event extraction is an information extraction task to obtain events from biomedical text, whose targets include the type, the trigger, and the respective arguments involved in an event. Traditional biomedical event extraction usually adopts a pipelined approach, which contains trigger identification, argument role recognition, and finally event construction either using specific rules or by machine learning. In this paper, we propose an n-ary relation extraction method based on the BERT pre-training model to construct Binding events, in order to capture the semantic information about an event's context and its participants. The experimental results show that our method achieves promising results on the GE11 and GE13 corpora of the BioNLP shared task with F1 scores of 63.14% and 59.40%, respectively. It demonstrates that by significantly improving the performance of Binding events, the overall performance of the pipelined event extraction approach or even exceeds those of current joint learning methods.

Attenuation of Polycyclic Aromatic Hydrocarbon (PAH)-Induced Carcinogenesis and Tumorigenesis by Omega-3 Fatty Acids in Mice In Vivo.

Lung cancer is the leading cause of cancer death worldwide. Polycyclic aromatic hydrocarbons (PAHs) are metabolized by the cytochrome P450 (CYP)1A and 1B1 to DNA-reactive metabolites, which could lead to mutations in critical genes, eventually resulting in cancer. Omega-3 fatty acids, such as eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA), are beneficial against cancers. In this investigation, we elucidated the mechanisms by which omega-3 fatty acids EPA and DHA will attenuate PAH-DNA adducts and lung carcinogenesis and tumorigenesis mediated by the PAHs BP and MC. Adult wild-type (WT) (A/J) mice, Cyp1a1-null, Cyp1a2-null, or Cyp1b1-null mice were exposed to PAHs benzo[a]pyrene (BP) or 3-methylcholanthrene (MC), and the effects of omega-3 fatty acid on PAH-mediated lung carcinogenesis and tumorigenesis were studied. The major findings were as follows: (i) omega-3 fatty acids significantly decreased PAH-DNA adducts in the lungs of each of the genotypes studied; (ii) decreases in PAH-DNA adduct levels by EPA/DHA was in part due to inhibition of CYP1B1; (iii) inhibition of soluble epoxide hydrolase (sEH) enhanced the EPA/DHA-mediated prevention of pulmonary carcinogenesis; and (iv) EPA/DHA attenuated PAH-mediated carcinogenesis in part by epigenetic mechanisms. Taken together, our results suggest that omega-3 fatty acids have the potential to be developed as cancer chemo-preventive agents in people.

Improving Room-Temperature Li-Metal Battery Performance by In Situ Creation of Fast Li+ Transport Pathways in a Polymer-Ceramic Electrolyte.

Composite polymer-ceramic electrolytes have shown considerable potential for high-energy-density Li-metal batteries as they combine the benefits of both polymers and ceramics. However, low ionic conductivity and poor contact with electrodes limit their practical usage. In this study, a highly conductive and stable composite electrolyte with a high ceramic loading is developed for high-energy-density Li-metal batteries. The electrolyte, produced through in situ polymerization and composed of a polymer called poly-1,3-dioxolane in a poly(vinylidene fluoride)/ceramic matrix, exhibits excellent room-temperature ionic conductivity of 1.2 mS cm-1 and high stability with Li metal over 1500 h. When tested in a Li|electrolyte|LiFePO4 battery, the electrolyte delivers excellent cycling performance and rate capability at room temperature, with a discharge capacity of 137 mAh g-1 over 500 cycles at 1 C. Furthermore, the electrolyte not only exhibits a high Li+ transference number of 0.76 but also significantly lowers contact resistance (from 157.8 to 2.1 Ω) relative to electrodes. When used in a battery with a high-voltage LiNi0.8 Mn0.1 Co0.1 O2 cathode, a discharge capacity of 140 mAh g-1 is achieved. These results show the potential of composite polymer-ceramic electrolytes in room-temperature solid-state Li-metal batteries and provide a strategy for designing highly conductive polymer-in-ceramic electrolytes with electrode-compatible interfaces.

Engineered Bacterial Outer Membrane Vesicles with Lipidated Heterologous Antigen as an Adjuvant-Free Vaccine Platform for Streptococcus suis.

Bacterial outer membrane vesicles (OMVs) are considered a promising vaccine platform for their high built-in adjuvanticity and ability to efficiently induce immune responses. OMVs can be engineered with heterologous antigens based on genetic engineering strategies. However, several critical issues should still be validated, including optimal exposure to the OMV surface, increased production of foreign antigens, nontoxicity, and induction of powerful immune protection. In this study, engineered OMVs with the lipoprotein transport machinery (Lpp) were designed to present SaoA antigen as a vaccine platform against Streptococcus suis. The results suggest that Lpp-SaoA fusions can be delivered on the OMV surface and do not have significant toxicity. Moreover, they can be engineered as lipoprotein and significantly accumulated in OMVs at high levels, thus accounting for nearly 10% of total OMV proteins. Immunization with OMVs containing Lpp-SaoA fusion antigen induced strong specific antibody responses and high levels of cytokines, as well as a balanced Th1/Th2 immune response. Furthermore, the decorated OMV vaccination significantly enhanced microbial clearance in a mouse infection model. It was found that antiserum against lipidated OMVs significantly promoted the opsonophagocytic uptake of S. suis in RAW246.7 macrophages. Lastly, OMVs engineered with Lpp-SaoA induced 100% protection against a challenge with 8× the 50% lethal dose (LD50) of S. suis serotype 2 and 80% protection against a challenge with 16× the LD50 in mice. Altogether, the results of this study provide a promising versatile strategy for the engineering of OMVs and suggest that Lpp-based OMVs may be a universal adjuvant-free vaccine platform for important pathogens. IMPORTANCE Bacterial outer membrane vesicles (OMVs) have become a promising vaccine platform due to their excellent built-in adjuvanticity properties. However, the location and amount of the expression of the heterologous antigen in the OMVs delivered by the genetic engineering strategies should be optimized. In this study, we exploited the lipoprotein transport pathway to engineer OMVs with heterologous antigen. Not only did lapidated heterologous antigen accumulate in the engineered OMV compartment at high levels, but also it was engineered to be delivered on the OMV surface, thus leading to the optimal activation of antigen-specific B cells and T cells. Immunization with engineered OMVs induced a strong antigen-specific antibodies in mice and conferred 100% protection against S. suis challenge. In general, the data of this study provide a versatile strategy for the engineering of OMVs and suggest that OMVs engineered with lipidated heterologous antigens may be a vaccine platform for significant pathogens.

Derivation and calibration of spectral response for a channeled spectropolarimeter.

The channeled spectropolarimeter (CSP) measures the spectrally-resolved Stokes vector from a snapshot by employing spectral modulation. The spectral modulation transfer function (SMTF) of the spectrometer preferentially suppresses the high-frequency channel amplitude in CSP, resulting in reduced measurement accuracy. This paper rigorously derives the SMTF theory and proposes an efficient calibration method for SMTF via channel shifting in a CSP. The SMTF value, obtained by channel shifting, is used to correct the high-frequency channel amplitude. Moreover, alignment and phase errors, as well as nonlinear dispersion, are compensated in situ. Other than rotating the retarder twice, no additional instruments or algorithms are required in the proposed method. In simulations and experiments, the proposed method shows high accuracy, with a maximum root-mean-square error (RMSE) of the reconstructed Stokes spectrum below 0.01, demonstrating its potential for enhancing the simplicity and practicability of Stokes CSP.

Joint learning-based causal relation extraction from biomedical literature.

Causal relation extraction of biomedical entities is one of the most complex tasks in biomedical text mining, which involves two kinds of information: entity relations and entity functions. One feasible approach is to take relation extraction and function detection as two independent sub-tasks. However, this separate learning method ignores the intrinsic correlation between them and leads to unsatisfactory performance. In this paper, we propose a joint learning model, which combines entity relation extraction and entity function detection to exploit their commonality and capture their inter-relationship, so as to improve the performance of biomedical causal relation extraction. Experimental results on the BioCreative-V Track 4 corpus show that our joint learning model outperforms the separate models in BEL statement extraction, achieving the F1 scores of 57.0% and 37.3% on the test set in Stage 2 and Stage 1 evaluations, respectively. This demonstrates that our joint learning system reaches the state-of-the-art performance in Stage 2 compared with other systems.

Recombinant-attenuated Salmonella enterica serovar Choleraesuis vector expressing the PlpE protein of Pasteurella multocida protects mice from lethal challenge.

BACKGROUND: Bacterial surface proteins play key roles in pathogenicity and often contribute to microbial adhesion and invasion. Pasteurella lipoprotein E (PlpE), a Pasteurella multocida (P. multocida) surface protein, has recently been identified as a potential vaccine candidate. Live attenuated Salmonella strains have a number of potential advantages as vaccine vectors, including immunization with live vector can mimic natural infections by organisms, lead to the induction of mucosal, humoral, and cellular immune responses. In this study, a previously constructed recombinant attenuated Salmonella Choleraesuis (S. Choleraesuis) vector rSC0016 was used to synthesize and secrete the surface protein PlpE of P. multocida to form the vaccine candidate rSC0016(pS-PlpE). Subsequently, the immunogenicity of S. Choleraesuis rSC0016(pS-PlpE) as an oral vaccine to induce protective immunity against P. multocida in mice was evaluated. RESULTS: After immunization, the recombinant attenuated S. Choleraesuis vector can efficiently delivered P. multocida PlpE protein in vivo and induced a specific immune response against this heterologous antigen in mice. In addition, compared with the inactivated vaccine, empty vector (rSC0016(pYA3493)) and PBS immunized groups, the rSC0016(pS-PlpE) vaccine candidate group induced higher antigen-specific mucosal, humoral and mixed Th1/Th2 cellular immune responses. After intraperitoneal challenge, the rSC0016(pS-PlpE) immunized group had a markedly enhanced survival rate (80%), a better protection efficiency than 60% of the inactivated vaccine group, and significantly reduced tissue damage. CONCLUSIONS: In conclusion, our study found that the rSC0016(pS-PlpE) vaccine candidate provided good protection against challenge with wild-type P. multocida serotype A in a mouse infection model, and may potentially be considered for use as a universal vaccine against multiple serotypes of P. multocida in livestock, including pigs.

Mueller matrix imaging polarimeter at the wavelength of 265 nm.

Mueller matrix imaging polarimeters (MMIPs) have been developed in the wavelength region of >400n m with great potential in many fields yet leaving a void of instrumentation and application in the ultraviolet (UV) region. For the first time to our knowledge, an UV-MMIP is developed for high resolution, sensitivity, and accuracy at the wavelength of 265 nm. A modified polarization state analyzer is designed and applied to suppress stray light for nice polarization images, and the errors of the measured Mueller matrices are calibrated to lower than 0.007 in pixel level. The finer performance of the UV-MMIP is demonstrated by the measurements of unstained cervical intraepithelial neoplasia (CIN) specimens. The contrasts of depolarization images obtained by the UV-MMIP are dramatically improved over those obtained by our previous VIS-MMIP at the wavelength of 650 nm. A distinct evolution of depolarization in normal cervical epithelium tissue, CIN-I, CIN-II, and CIN-III specimens can be observed by the UV-MMIP with mean depolarization promotion by up to 20 times. This evolution could provide important evidence for CIN staging but can hardly be distinguished by the VIS-MMIP. The results prove that the UV-MMIP could be an effective tool in polarimetric applications with higher sensitivity.

Multi-probe attention neural network for COVID-19 semantic indexing.

BACKGROUND: The COVID-19 pandemic has increasingly accelerated the publication pace of scientific literature. How to efficiently curate and index this large amount of biomedical literature under the current crisis is of great importance. Previous literature indexing is mainly performed by human experts using Medical Subject Headings (MeSH), which is labor-intensive and time-consuming. Therefore, to alleviate the expensive time consumption and monetary cost, there is an urgent need for automatic semantic indexing technologies for the emerging COVID-19 domain. RESULTS: In this research, to investigate the semantic indexing problem for COVID-19, we first construct the new COVID-19 Semantic Indexing dataset, which consists of more than 80 thousand biomedical articles. We then propose a novel semantic indexing framework based on the multi-probe attention neural network (MPANN) to address the COVID-19 semantic indexing problem. Specifically, we employ a k-nearest neighbour based MeSH masking approach to generate candidate topic terms for each input article. We encode and feed the selected candidate terms as well as other contextual information as probes into the downstream attention-based neural network. Each semantic probe carries specific aspects of biomedical knowledge and provides informatively discriminative features for the input article. After extracting the semantic features at both term-level and document-level through the attention-based neural network, MPANN adopts a linear multi-view classifier to conduct the final topic prediction for COVID-19 semantic indexing. CONCLUSION: The experimental results suggest that MPANN promises to represent the semantic features of biomedical texts and is effective in predicting semantic topics for COVID-19 related biomedical articles.

Reconstruction and calibration methods for a Mueller channeled spectropolarimeter.

Channeled spectropolarimeter (CSP) measures spectrally resolved Stokes vector of light and Mueller matrix of sample from a snapshot. While reconstruction and calibration methods for Stokes CSP have been well established, their Mueller CSP counterparts are lacking. In this paper, we propose methods for Mueller spectrum reconstruction and Mueller CSP calibration. Mueller CSP is modeled as a modulation matrix, linking the Mueller spectrum to be measured and the modulated spectrum from the spectrometer. We describe an optimization problem to solve the Mueller spectrum, where both the regularizer and the residual threshold constrain the result, making our reconstruction accurate, efficient, and noise-robust. The Stokes spectrum generated by polarization state generator and the analyzing vector of polarization state analyzer are measured in situ, the convolution of which construct the calibrated modulation matrix of Mueller CSP. Total polarimetric errors and spectroscopic errors are treated as a whole and represented by the calibrated modulation matrix. Both imaging and non-imaging Mueller CSP are experimentally calibrated. Reconstruction results show high accuracy with a root-mean-square error (RMSE) of 0.0371. The proposed methods help make Mueller CSP practical and have the potential to be general reconstruction and calibration methods for imaging and non-imaging Stokes-Mueller CSP.

Optimized spatially modulated polarimetry with an efficient calibration method and hybrid gradient descent reconstruction.

High accuracy and fast polarization measurements at a low light field are significant in various applications, spanning from quantum optics to diagnosis of living biological tissue. In this paper, we developed an optimized spatially modulated polarimetry (OSMP) with an efficient calibration method that establishes a quantitative link between the intensity distribution of an arbitrary incident polarization state and four intensity distributions of specific input polarization states. Such a calibration method not only considers the total polarimetric errors induced by polarization elements and the focusing lens but also simplifies the procedure of calibration. A hybrid gradient descent (HGD) algorithm, combining the rapidity of optimization of gradient descent (GD) algorithm and the accuracy of optimization of direct enumeration (DE) algorithm, was proposed to restructure the Stokes parameters. Experiment results illustrate that the proposed method can significantly improve the speed and accuracy of polarization measurements over existing spatially modulated polarimeters based on the vortex wave retarder, whether in strong or low light fields.

Optimal design for a broadband Stokes polarimeter of liquid crystal variable retarders.

Liquid crystal variable retarders (LCVRs) are the core component for rapid and high-precision broadband polarization detection. Additionally, the ability to suppress noise greatly affects the results of polarization measurements. In this work, a solving optimal design approach is proposed for building a high-performance broadband Stokes polarimeter based on LCVRs, which greatly reduces the influences of data fluctuation from liquid crystals and dispersion on the experimental results. This method relies on evaluation criteria of the condition number (CN) to build a gradual optimization that includes the following three steps: fixing the fast axis angles, meeting the requirements of a wideband, and ensuring a minimum CN. Additionally, with the method of increasing the measurement analysis vector, we ensure the whole band in the low CN and offer a solution to the problem of the difficulty in optimizing the LCVRs caused by the large change of retardance at 490-700 nm. Finally, the rapid and high-precision Stokes measurement of 490-700 nm wavelengths is achieved. We test the performance of the polarimeter after optimization in our simulation and experiment, which shows that the total RMS error is less than 0.032 and the single point error is small. This work not only reduces the influence of LCVR error on the experimental results but also makes it possible to apply LCVRs to 490-700 nm detection.

Holistic and efficient calibration method for Mueller matrix imaging polarimeter with a high numerical aperture.

High-numerical aperture (N A>0.6) Mueller matrix imaging polarimeter (MMIP) (high-NA MMIP) is urgently needed for higher resolution. Usually, the working distance of high-NA MMIP is too short to perform in situ calibration by a usual reference sample, such as polarizer and retarder plates. The polarization effects of the substrate that attach the sample are never calibrated. So, the resolution and accuracy of the MMIP is hard to further promote. In this paper, a holistic and efficient calibration method is innovated for high-NA MMIP. Two film polarizers and a film retarder as well as a blank substrate are first adopted as the reference samples in calibration. Different from the conventional eigenvalue calibration method (ECM), the holistic calibration theory and process are established. All polarimetric errors arising from the devices, subsystems, and the substrate can be calibrated in one process. The normalized measurement error is less than 0.0024 for NA 0.95 MMIP, which is an order of magnitude lower than those of NA 0.1 and 0.2 MMIPs in publications. The excellent performance of calibrated high-NA MMIP is demonstrated by tissue polarimetry with higher resolution, accuracy, and more appropriate dynamic range.

Efficient calibration method of total polarimetric errors in a channeled spectropolarimeter.

An efficient calibration method of total polarimetric errors in a channeled spectropolarimeter (CSP) is proposed and experimentally demonstrated. Total polarimetric errors, including alignment and retardance errors as well as those caused by nonideal retarders and the polarizer in CSP, are considered and calibrated. We first construct the calibrated modulation matrix of CSP by directly measuring the Mueller matrix spectrum of the polarization module in CSP. Compared to previously reported calibration works that required 1074 measurements, our calibration requires only 16 individual measurements, which reduces the measurement time by 67-fold while ensuring high accuracy with a maximum rms error less than 0.02. Further experimental test on three types of different CSP systems confirms the efficiency, reliability, and accuracy of the proposed calibration method.

Observation of Strong J-Aggregate Light Emission in Monolayer Molecular Crystal on Hexagonal Boron Nitride.

J-aggregates are widely used in studies of light-matter interaction and organic optoelectronic devices. Although J-aggregate films can be fabricated on salt by epitaxial growth method, the size is limited to hundreds of nanometer. In this work, with hexagonal boron nitride (h-BN) as a substrate, highly crystalline J-aggregate ultrathin films of N,N'-ditridecylperylene 3,4,9,10-tetracarboxylic diimide (PTCDI-C13) are achieved by physical vapor transport (PVT) method. Significant bathochromically shifted absorption band and narrowed 0-0 transition are observed in the monolayer PTCDI-C13 crystal on h-BN. The exciton coherence number Ncoh of monolayer J-aggregate film extracted from the photoluminescence (PL) spectrum is up to 15 at T = 140 K, which is higher than that of the epitaxially grown layer on salt. Beyond the first molecular layer, the multilayer crystal on h-BN is dominated by H-aggregates. Further study shows that that the first molecular layer on h-BN adopts the highly ordered face-on configuration, while the overlayers adopt the edge-on motif. As a comparison, only H-aggregate PTCDI-C13 ultrathin films are found on SiO2 substrates, but no J-aggregates. The results suggest that high-quality J-aggregates can be prepared by utilizing appropriate substrates via physical vapor transport.

Ganoderma Lucidum Polysaccharide, an Extract from Ganoderma Lucidum, Exerts Suppressive Effect on Cervical Cancer Cell Malignancy through Mitigating Epithelial-Mesenchymal and JAK/STAT5 Signaling Pathway.

BACKGROUND/AIMS: We aimed to explore whether ganoderma lucidum polysaccharide (GLP) exhibits antitumor effect on cervical cancer cells. METHODS AND RESULTS: Different concentration of GLP was used to treat cervical cell. The data from cell counting kit-8 assay proved that the optimal working concentration and time of GLP were 200 µg/mL and treated for 48 h. The transwell assay demonstrated that GLP could attenuate the invasion and migration abilities of cervical cancer cells. Moreover, flow cytometry illustrated that GLP could promote the apoptosis of cervical cancer cells and limit the cycle of cervical cancer cells. Western blot assay discovered that the expression of proapoptosis proteins including Bax, Cleaved Caspases 3 and 9 increased and the antiapoptosis protein Bcl-2 decreased after treated with GLP. Moreover, we found that the expression of E-cadherin was increased, and N-cadherin, Vimentin, and Slug were decreased. Meanwhile, the expression of phosphorylated-JAK and phosphorylated-STAT5 was also decreased in cervical cancer cells treated by GLP, suggesting the inhibitory effect on JAK/STAT5 pathways. CONCLUSIONS: All of these data illustrated that GLP could alleviate the activity and aggressiveness, block the cell cycle, and promote the apoptosis of cervical cancer cells, which were possible via inhibiting epithelial-mesenchymal and JAK/STAT5 pathways.

Chemical-induced disease relation extraction via attention-based distant supervision.

BACKGROUND: Automatically understanding chemical-disease relations (CDRs) is crucial in various areas of biomedical research and health care. Supervised machine learning provides a feasible solution to automatically extract relations between biomedical entities from scientific literature, its success, however, heavily depends on large-scale biomedical corpora manually annotated with intensive labor and tremendous investment. RESULTS: We present an attention-based distant supervision paradigm for the BioCreative-V CDR extraction task. Training examples at both intra- and inter-sentence levels are generated automatically from the Comparative Toxicogenomics Database (CTD) without any human intervention. An attention-based neural network and a stacked auto-encoder network are applied respectively to induce learning models and extract relations at both levels. After merging the results of both levels, the document-level CDRs can be finally extracted. It achieves the precision/recall/F1-score of 60.3%/73.8%/66.4%, outperforming the state-of-the-art supervised learning systems without using any annotated corpus. CONCLUSION: Our experiments demonstrate that distant supervision is promising for extracting chemical disease relations from biomedical literature, and capturing both local and global attention features simultaneously is effective in attention-based distantly supervised learning.

Association between elevated placental polycyclic aromatic hydrocarbons (PAHs) and PAH-DNA adducts from Superfund sites in Harris County, and increased risk of preterm birth (PTB).

The preterm birth (PTB) rate in Harris County, Texas, exceeds the U.S. rate (11.4% vs.9.6%), and there are 15 active Superfund sites in Harris County. Polycyclic aromatic hydrocarbons (PAHs) are contaminants of concern (COC) at Superfund sites across the nation. In this investigation, we tested the hypothesis that higher levels of exposure to PAHs and PAH-DNA adducts in placenta of women living near Superfund sites contribute to the increased rate of PTBs. Levels of benzo[a]pyene (BP), benzo[b]fluorene (BbF) and dibenz[a,h]anthracene (DBA), were higher in placentae from preterm deliveries compared with term deliveries in women living near Superfund sites, whereas this was not the case for women living in non-Superfund site areas. Among the PAHs, DBA levels were significantly higher than BP or BbF, and DBA levels were inversely correlated with gestational age at delivery and birth weight. Bulky PAH-DNA adducts are more prevalent in placental tissue from individuals residing near Superfund sites. Expression of Ah receptor (AHR) and NF-E2-related factor 2 (NRF2) was decreased in preterm deliveries in subjects residing near Superfund sites. Unbiased metabolomics revealed alterations in pathways involved in pentose phosphate, inositol phosphate and starch and sucrose metabolism in preterm subjects in Superfund site areas. In summary, this is the first report showing an association between PAH levels, DNA adducts, and modulation of endogenous metabolic pathways with PTBs in subjects residing near Superfund sites, and further studies could lead to novel strategies in the understanding of the mechanisms by which PAHs contribute to PTBs in women.

Hierarchical sequence labeling for extracting BEL statements from biomedical literature.

BACKGROUND: Extracting relations between bio-entities from biomedical literature is often a challenging task and also an essential step towards biomedical knowledge expansion. The BioCreative community has organized a shared task to evaluate the robustness of the causal relationship extraction algorithms in Biological Expression Language (BEL) from biomedical literature. METHOD: We first map the sentence-level BEL statements in the BC-V training corpus to the corresponding text segments, thus generating hierarchically tagged training instances. A hierarchical sequence labeling model was afterwards induced from these training instances and applied to the test sentences in order to construct the BEL statements. RESULTS: The experimental results on extracting BEL statements from BioCreative V Track 4 test corpus show that our method achieves promising performance with an overall F-measure of 31.6%. Furthermore, it has the potential to be enhanced by adopting more advanced machine learning approaches. CONCLUSION: We propose a framework for hierarchical relation extraction using hierarchical sequence labeling on the instance-level training corpus derived from the original sentence-level corpus via word alignment. Its main advantage is that we can make full use of the original training corpus to induce the sequence labelers and then apply them to the test corpus.

High-performance chemical vapor deposited graphene-on-silicon nitride waveguide photodetectors.

Waveguide photodetectors integrated with graphene have demonstrated potential for ultrafast response and broadband operation. Here, we demonstrate high-performance chemical vapor deposited graphene-on-silicon nitride waveguide photodetectors by enhancing the absorption of light propagating in the transverse-magnetic mode through a metal-graphene junction. A doubling in responsivity is experimentally observed. In our zero-biased metal-graphene junction, a 15 mA W-1 intrinsic responsivity and a 30 GHz bandwidth are achieved at ∼1550 nm. The results are comparable to those obtained from the best pristine graphene-based photodetectors. Our work enables new architectures for high-performance optoelectronic devices based on the graphene-on-silicon nitride platform.