Abstract: Deep nets generalize well despite having more parameters than the number of training samples. Recent works try to give an explanation using PAC-Bayes and Margin-based analyses, but do not as yet result in sample complexity bounds better than naive parameter counting. The current paper shows generalization bounds that're orders of magnitude better in practice. These rely upon new succinct reparametrizations of the trained net --- a compression that is explicit and efficient. These yield generalization bounds via a simple compression-based framework introduced here. Our results also provide some theoretical justification for widespread empirical success in compressing deep nets. Analysis of correctness of our compression relies upon some newly identified \textquotedblleft noise stability\textquotedblright properties of trained deep nets, which are also experimentally verified. The study of these properties and resulting generalization bounds are also extended to convolutional nets, which had eluded earlier attempts on proving generalization.

Abstract: The efficacy of disease resistance genes in plants decreases over time because of the selection of virulent pathogen genotypes. A key goal of crop protection programs is to increase the durability of the resistance conferred by these genes. The spatial and temporal deployment of plant disease resistance genes is considered to be a major factor determining their durability. In the literature, four principal strategies combining resistance genes over time and space have been considered to delay the evolution of virulent pathogen genotypes: cultivars mixture, rotation, landscape deployment, pyramiding. We reviewed this literature with the aim of determining which deployment strategy results in the greatest durability of resistance genes. Although theoretical and empirical studies comparing deployment strategies of more than one resistance gene are very scarce, they suggest that the overall durability of disease resistance genes can be increased by combining their presence in the same plant (pyramiding). Retrospective analyses of field monitoring data also suggest that the pyramiding of disease resistance genes within a plant is the most durable strategy. By extension, we suggest that the combination of disease resistance genes with other practices for pathogen control (pesticides, farming practices) may be a relevant management strategy to slow down the evolution of virulent pathogen genotypes.

Abstract: Traffic-related sources have been recognized as a significant contributor of particulate matter particularly within major cities. Exhaust and non-exhaust traffic-related sources are estimated to contribute almost equally to traffic-related PM10 emissions. Non-exhaust particles can be generated either from non-exhaust sources such as brake, tyre, clutch and road surface wear or already exist in the form of deposited material at the roadside and become resuspended due to traffic-induced turbulence. Among non-exhaust sources, brake wear can be a significant particulate matter (PM) contributor, particularly within areas with high traffic density and braking frequency. Studies mention that in urban environments, brake wear can contribute up to 55 % by mass to total non-exhaust traffic-related PM10 emissions and up to 21 % by mass to total traffic-related PM10 emissions, while in freeways, this contribution is lower due to lower braking frequency. As exhaust emissions control become stricter, relative contributions of non-exhaust sources—and therefore brake wear—to traffic-related emissions will become more significant and will raise discussions on possible regulatory needs. The aim of the present literature review study is to present the state-of-the-art of the different aspects regarding PM resulting from brake wear and provide all the necessary information in terms of importance, physicochemical characteristics, emission factors and possible health effects.

Abstract: In the early days, content-based image retrieval (CBIR) was studied with global features. Since 2003, image retrieval based on local descriptors ( de facto SIFT) has been extensively studied for over a decade due to the advantage of SIFT in dealing with image transformations. Recently, image representations based on the convolutional neural network (CNN) have attracted increasing interest in the community and demonstrated impressive performance. Given this time of rapid evolution, this article provides a comprehensive survey of instance retrieval over the last decade. Two broad categories, SIFT-based and CNN-based methods, are presented. For the former, according to the codebook size, we organize the literature into using large/medium-sized/small codebooks. For the latter, we discuss three lines of methods, i.e., using pre-trained or fine-tuned CNN models, and hybrid methods. The first two perform a single-pass of an image to the network, while the last category employs a patch-based feature extraction scheme. This survey presents milestones in modern instance retrieval, reviews a broad selection of previous works in different categories, and provides insights on the connection between SIFT and CNN-based methods. After analyzing and comparing retrieval performance of different categories on several datasets, we discuss promising directions towards generic and specialized instance retrieval.

Abstract: The fact that various immune cells, including macrophages, can be found in tumor tissues has long been known. With the introduction of concept that macrophages differentiate into a classically or alternatively activated phenotype, the role of tumor-associated macrophages (TAMs) is now beginning to be elucidated. TAMs act as “protumoral macrophages,” contributing to disease progression. TAMs can promote initiation and metastasis of tumor cells, inhibit antitumor immune responses mediated by T cells, and stimulate tumor angiogenesis and subsequently tumor progression. As the relationship between TAMs and malignant tumors becomes clearer, TAMs are beginning to be seen as potential biomarkers for diagnosis and prognosis of cancers, as well as therapeutic targets in these cases. In this review, we will discuss the origin, polarization, and role of TAMs in human malignant tumors, as well as how TAMs can be used as diagnostic and prognostic biomarkers and therapeutic targets of cancer in clinics.

Abstract: The absence of motor vehicle traffic and suspended manufacturing during the coronavirus disease 2019 (COVID-19) pandemic in China enabled assessment of the efficiency of air pollution mitigation. Up to 90% reduction of certain emissions during the city-lockdown period can be identified from satellite and ground-based observations. Unexpectedly, extreme particulate matter levels simultaneously occurred in northern China. Our synergistic observation analyses and model simulations show that anomalously high humidity promoted aerosol heterogeneous chemistry, along with stagnant airflow and uninterrupted emissions from power plants and petrochemical facilities, contributing to severe haze formation. Also, because of nonlinear production chemistry and titration of ozone in winter, reduced nitrogen oxides resulted in ozone enhancement in urban areas, further increasing the atmospheric oxidizing capacity and facilitating secondary aerosol formation.

Abstract: Adversarial training, a method for learning robust deep networks, is typically assumed to be more expensive than traditional training due to the necessity of constructing adversarial examples via a first-order method like projected gradient decent (PGD). In this paper, we make the surprising discovery that it is possible to train empirically robust models using a much weaker and cheaper adversary, an approach that was previously believed to be ineffective, rendering the method no more costly than standard training in practice. Specifically, we show that adversarial training with the fast gradient sign method (FGSM), when combined with random initialization, is as effective as PGD-based training but has significantly lower cost. Furthermore we show that FGSM adversarial training can be further accelerated by using standard techniques for efficient training of deep networks, allowing us to learn a robust CIFAR10 classifier with 45% robust accuracy at epsilon=8/255 in 6 minutes, and a robust ImageNet classifier with 43% robust accuracy at epsilon=2/255 in 12 hours, in comparison to past work based on ``free'' adversarial training which took 10 and 50 hours to reach the same respective thresholds.

Abstract: The COVID-19 pandemic quickly led to the closure of universities and colleges around the world, in hopes that public health officials’ advice of social distancing could help to flatten the infectio...

Abstract: This study reports on the prevalence, intensity, and timing of an altered sense of smell or taste in patients with SARS-CoV-2 infections.

Abstract: There is no doubt that ionic liquids have become a major subject of study for modern chemistry. We have become used to ever more publications in the field each year, although there is some evidence that this is beginning to plateau at approximately 3500 papers each year. They have been the subject of several major reviews and books, dealing with different applications and aspects of their behaviours. In this article, I will show a little of how interest in ionic liquids grew and developed.

Abstract: The recent Ebola virus (EBOV) outbreak in West Africa was the largest recorded in history with over 28,000 cases, resulting in >11,000 deaths including >500 healthcare workers. A focused screening and lead optimization effort identified 4b (GS-5734) with anti-EBOV EC50 = 86 nM in macrophages as the clinical candidate. Structure activity relationships established that the 1′-CN group and C-linked nucleobase were critical for optimal anti-EBOV potency and selectivity against host polymerases. A robust diastereoselective synthesis provided sufficient quantities of 4b to enable preclinical efficacy in a non-human-primate EBOV challenge model. Once-daily 10 mg/kg iv treatment on days 3–14 postinfection had a significant effect on viremia and mortality, resulting in 100% survival of infected treated animals [Nature 2016, 531, 381−385]. A phase 2 study (PREVAIL IV) is currently enrolling and will evaluate the effect of 4b on viral shedding from sanctuary sites in EBOV survivors.

Abstract: Here we describe a protocol for advanced CUBIC (Clear, Unobstructed Brain/Body Imaging Cocktails and Computational analysis). The CUBIC protocol enables simple and efficient organ clearing, rapid imaging by light-sheet microscopy and quantitative imaging analysis of multiple samples. The organ or body is cleared by immersion for 1-14 d, with the exact time required dependent on the sample type and the experimental purposes. A single imaging set can be completed in 30-60 min. Image processing and analysis can take <1 d, but it is dependent on the number of samples in the data set. The CUBIC clearing protocol can process multiple samples simultaneously. We previously used CUBIC to image whole-brain neural activities at single-cell resolution using Arc-dVenus transgenic (Tg) mice. CUBIC informatics calculated the Venus signal subtraction, comparing different brains at a whole-organ scale. These protocols provide a platform for organism-level systems biology by comprehensively detecting cells in a whole organ or body.

Abstract: This paper proposes the novel Pose Guided Person Generation Network (PG$^2$) that allows to synthesize person images in arbitrary poses, based on an image of that person and a novel pose. Our generation framework PG$^2$ utilizes the pose information explicitly and consists of two key stages: pose integration and image refinement. In the first stage the condition image and the target pose are fed into a U-Net-like network to generate an initial but coarse image of the person with the target pose. The second stage then refines the initial and blurry result by training a U-Net-like generator in an adversarial way. Extensive experimental results on both 128$\times$64 re-identification images and 256$\times$256 fashion photos show that our model generates high-quality person images with convincing details.

Abstract: Understanding cell-fate decisions during tumorigenesis and metastasis is a major challenge in modern cancer biology. One canonical cell-fate decision that cancer cells undergo is Epithelial-to-Mesenchymal Transition (EMT) and its reverse Mesenchymal-to-Epithelial Transition (MET). While transitioning between these two phenotypes – epithelial and mesenchymal, cells can also attain a hybrid epithelial/ mesenchymal (i.e. partial or intermediate EMT) phenotype. Cells in this phenotype have mixed epithelial (eg. adhesion) and mesenchymal (eg. migration) properties, thereby allowing them to move collectively as clusters of Circulating Tumor Cells (CTCs). If these clusters enter the circulation, they can be more apoptosis-resistant and more capable of initiating metastatic lesions than cancer cells moving individually with wholly mesenchymal phenotypes, having undergone a complete EMT. Here, we review the operating principles of the core regulatory network for EMT/MET that acts as a ‘three-way’ switch giving rise to three distinct phenotypes – epithelial, mesenchymal and hybrid epithelial/mesenchymal. We further characterize this hybrid E/M phenotype in terms of its capabilities in terms of collective cell migration, tumor-initiation, cell-cell communication, and drug resistance. We elucidate how the highly interconnected coupling between these modules coordinates cell-fate decisions among a population of cancer cells in the dynamic tumor, hence facilitating tumor-stroma interactions, formation of CTC clusters, and consequently cancer metastasis. Finally, we discuss the multiple advantages that the hybrid epithelial/mesenchymal phenotype have as compared to a complete EMT phenotype and argue that these collectively migrating cells are the primary ‘bad actors’ of metastasis.

Abstract: . This paper presents an update and extension of HYDE, the History Database of the Global Environment (HYDE version 3.2). HYDE is an internally consistent combination of historical population estimates and allocation algorithms with time-dependent weighting maps for land use. Categories include cropland, with new distinctions for irrigated and rain-fed crops (other than rice) and irrigated and rain-fed rice. Grazing lands are also provided, divided into more intensively used pasture and less intensively used rangeland, and further specified with respect to conversion of natural vegetation to facilitate global change modellers. Population is represented by maps of total, urban, rural population, population density and built-up area. The period covered is 10 000 before Common Era (BCE) to 2015 Common Era (CE). All data can be downloaded from https://doi.org/10.17026/dans-25g-gez3 . We estimate that global population increased from 4.4 million people (we also estimate a lower range Cropland occupied approximately less than 1 % of the global land area (13 037 Mha, excluding Antarctica) for a long time period until 1 CE, quite similar to the grazing land area. In the following centuries the share of global cropland slowly grew to 2.2 % in 1700 CE (ca. 293 Mha, uncertainty range 220–367 Mha), 4.4 % in 1850 CE (578 Mha, range 522–637 Mha) and 12.2 % in 2015 CE (ca. 1591 Mha, range 1572–1604 Mha). Cropland can be further divided into rain-fed and irrigated land, and these categories can be further separated into rice and non-rice. Rain-fed croplands were much more common, with 2.2 % in 1700 CE (289 Mha, range 217–361 Mha), 4.2 % (549 Mha, range 496–606 Mha) in 1850 CE and 10.1 % (1316 Mha, range 1298–1325 Mha) in 2015 CE, while irrigated croplands used less than 0.05 % (4.3 Mha, range 3.1–5.5 Mha), 0.2 % (28 Mha, range 25–31 Mha) and 2.1 % (277 Mha, range 273–278 Mha) in 1700, 1850 and 2015 CE, respectively. We estimate the irrigated rice area (paddy) to be 0.1 % (13 Mha, range 9–16 Mha) in 1700 CE, 0.2 % (28 Mha, range 26–31 Mha) in 1850 CE and 0.9 % (118 Mha, range 117–120 Mha) in 2015 CE. The estimates for land used for grazing are much more uncertain. We estimate that the share of grazing land grew from 5.1 % in 1700 CE (667 Mha, range 507–820 Mha) to 9.6 % in 1850 CE (1192 Mha, range 1068–1304 Mha) and 24.9 % in 2015 CE (3241 Mha, range 3211–3270 Mha). To aid the modelling community we have divided land used for grazing into more intensively used pasture, less intensively used converted rangeland and less or unmanaged natural unconverted rangeland. Pasture occupied 1.1 % in 1700 CE (145 Mha, range 79–175 Mha), 1.9 % in 1850 CE (253 Mha, range 218–287 Mha) and 6.0 % (787 Mha, range 779–795 Mha) in 2015 CE, while rangelands usually occupied more space due to their occurrence in more arid regions and thus lower yields to sustain livestock. We estimate converted rangeland at 0.6 % in 1700 CE (82 Mha range 66–93 Mha), 1 % in 1850 CE (129 Mha range 118–136 Mha) and 2.4 % in 2015 CE (310 Mha range 306–312 Mha), while the unconverted natural rangelands occupied approximately 3.4 % in 1700 CE (437 Mha, range 334–533 Mha), 6.2 % in 1850 CE (810 Mha, range 733–881 Mha) and 16.5 % in 2015 CE (2145 Mha, range 2126–2164 Mha).

Abstract: This article reports world averages of measurements of b-hadron, c-hadron, and tau-lepton properties obtained by the Heavy Flavor Averaging Group (HFAG) using results available through the end of 2011. In some cases results available in the early part of 2012 are included. For the averaging, common input parameters used in the various analyses are adjusted (rescaled) to common values, and known correlations are taken into account. The averages include branching fractions, lifetimes, neutral meson mixing parameters, CP violation parameters, parameters of semileptonic decays and CKM matrix elements.

Abstract: Remotely sensed data can reinforce the abilities of water resources researchers and decision makers to monitor waterbodies more effectively. Remote sensing techniques have been widely used to measure the qualitative parameters of waterbodies (i.e., suspended sediments, colored dissolved organic matter (CDOM), chlorophyll-a, and pollutants). A large number of different sensors on board various satellites and other platforms, such as airplanes, are currently used to measure the amount of radiation at different wavelengths reflected from the water’s surface. In this review paper, various properties (spectral, spatial and temporal, etc.) of the more commonly employed spaceborne and airborne sensors are tabulated to be used as a sensor selection guide. Furthermore, this paper investigates the commonly used approaches and sensors employed in evaluating and quantifying the eleven water quality parameters. The parameters include: chlorophyll-a (chl-a), colored dissolved organic matters (CDOM), Secchi disk depth (SDD), turbidity, total suspended sediments (TSS), water temperature (WT), total phosphorus (TP), sea surface salinity (SSS), dissolved oxygen (DO), biochemical oxygen demand (BOD) and chemical oxygen demand (COD).

Abstract: Neural network training relies on our ability to find "good" minimizers of highly non-convex loss functions. It is well known that certain network architecture designs (e.g., skip connections) produce loss functions that train easier, and well-chosen training parameters (batch size, learning rate, optimizer) produce minimizers that generalize better. However, the reasons for these differences, and their effect on the underlying loss landscape, is not well understood. In this paper, we explore the structure of neural loss functions, and the effect of loss landscapes on generalization, using a range of visualization methods. First, we introduce a simple "filter normalization" method that helps us visualize loss function curvature, and make meaningful side-by-side comparisons between loss functions. Then, using a variety of visualizations, we explore how network architecture affects the loss landscape, and how training parameters affect the shape of minimizers.

Abstract: Universal fault-tolerant quantum computers will require error-free execution of long sequences of quantum gate operations, which is expected to involve millions of physical qubits. Before the full power of such machines will be available, near-term quantum devices will provide several hundred qubits and limited error correction. Still, there is a realistic prospect to run useful algorithms within the limited circuit depth of such devices. Particularly promising are optimization algorithms that follow a hybrid approach: the aim is to steer a highly entangled state on a quantum system to a target state that minimizes a cost function via variation of some gate parameters. This variational approach can be used both for classical optimization problems as well as for problems in quantum chemistry. The challenge is to converge to the target state given the limited coherence time and connectivity of the qubits. In this context, the quantum volume as a metric to compare the power of near-term quantum devices is discussed. With focus on chemistry applications, a general description of variational algorithms is provided and the mapping from fermions to qubits is explained. Coupled-cluster and heuristic trial wave-functions are considered for efficiently finding molecular ground states. Furthermore, simple error-mitigation schemes are introduced that could improve the accuracy of determining ground-state energies. Advancing these techniques may lead to near-term demonstrations of useful quantum computation with systems containing several hundred qubits.

Abstract: Letters RESEARCH LETTER Trends in Management for Patients With Localized Prostate Cancer, 1990-2013 A growing literature supports the safety and efficacy of active surveillance for patients with low-risk prostate cancer. How- ever, the experience behind this literature is based almost en- tirely in academic centers, and prior reports have consis- tently found surveillance generally underused in most other settings. 1,2 Conversely, high-risk tumors have been under- treated with androgen deprivation treatment alone. 2,3 Re- cent trends in community-based practice patterns have not been well documented. Methods | Cancer of the Prostate Strategic Urologic Research En- deavor (CaPSURE) is a national registry accruing men with pros- tate cancer diagnosed at 45 urology practices across the United States since 1995. A mix of large and small practices are in- cluded. All but 3 are community-based practices and 28 states across all regions are represented. Both prospective enroll- ment of newly diagnosed men and retrospective enrollment of previously diagnosed men were permitted before 1998; how- ever, since 1998 all enrollment has been prospective. Approximately 90% of eligible patients are accrued. Urologists report clinical data; patients provide written consent under central institutional review board supervi- sion. Other methodological details have been reported. 4 We analyzed men with tumors classified as stage cT3aN0M0 or lower managed with prostatectomy, radiation, androgen deprivation monotherapy, or active surveillance/watchful waiting between 1990 and 2013. Only recently have these 2 terms been clearly separated, 1 and CaPSURE has histori- cally recorded them as a single category. There were 656 men (5.9%) with missing treatment data who were excluded. Cancer risk was stratified using the validated Cancer of the Prostate Risk Assessment (CAPRA) score. 5 We analyzed treatment trends over 5-year intervals in the full cohort and in a subset of men aged 75 years or older. We calculated Mantel-Haenszel tests for trends over time. There have been changes in the CaPSURE sites over time (eg, some have closed or withdrawn and others have been added). A subset analysis including only practices steadily contributing patients found substantially similar results. Analyses were performed with Stata version 12.1 (StataCorp). Statistical tests were 2-tailed with α = .05. Figure 1. Treatment Trends for the Overall Cohort in the Cancer of the Prostate Strategic Urologic Research Endeavor (CaPSURE) Registry Type of treatment Active surveillance/watchful waiting Radical prostatectomy CAPRA score range, 3-5 CAPRA score range, 6-10 Patients With Prostate Cancer, % Patients With Prostate Cancer, % Patients With Prostate Cancer, % CAPRA score range, 0-2 Year of Diagnosis Radiation therapy Primary androgen deprivation therapy Year of Diagnosis Year of Diagnosis Error bars indicate 95% confidence intervals; CAPRA, Cancer of the Prostate Risk Assessment. JAMA July 7, 2015 Volume 314, Number 1 (Reprinted) Copyright 2015 American Medical Association. All rights reserved. Downloaded From: http://jama.jamanetwork.com/ by a UCSF LIBRARY User on 07/28/2016 jama.com

Abstract: Describing the color and textural information of a person image is one of the most crucial aspects of person re-identification. In this paper, we present a novel descriptor based on a hierarchical distribution of pixel features. A hierarchical covariance descriptor has been successfully applied for image classification. However, the mean information of pixel features, which is absent in covariance, tends to be major discriminative information of person images. To solve this problem, we describe a local region in an image via hierarchical Gaussian distribution in which both means and covariances are included in their parameters. More specifically, we model the region as a set of multiple Gaussian distributions in which each Gaussian represents the appearance of a local patch. The characteristics of the set of Gaussians are again described by another Gaussian distribution. In both steps, unlike the hierarchical covariance descriptor, the proposed descriptor can model both the mean and the covariance information of pixel features properly. The results of experiments conducted on five databases indicate that the proposed descriptor exhibits re-markably high performance which outperforms the state-of-the-art descriptors for person re-identification.

Abstract: Recently pre-trained models have achieved state-of-the-art results in various language understanding tasks. Current pre-training procedures usually focus on training the model with several simple tasks to grasp the co-occurrence of words or sentences. However, besides co-occurring information, there exists other valuable lexical, syntactic and semantic information in training corpora, such as named entities, semantic closeness and discourse relations. In order to extract the lexical, syntactic and semantic information from training corpora, we propose a continual pre-training framework named ERNIE 2.0 which incrementally builds pre-training tasks and then learn pre-trained models on these constructed tasks via continual multi-task learning. Based on this framework, we construct several tasks and train the ERNIE 2.0 model to capture lexical, syntactic and semantic aspects of information in the training data. Experimental results demonstrate that ERNIE 2.0 model outperforms BERT and XLNet on 16 tasks including English tasks on GLUE benchmarks and several similar tasks in Chinese. The source codes and pre-trained models have been released at https://github.com/PaddlePaddle/ERNIE.

Abstract: The structure and function of biomolecules are strongly influenced by their hydration shells. Structural fluctuations and molecular excitations of hydrating water molecules cover a broad range in space and time, from individual water molecules to larger pools and from femtosecond to microsecond time scales. Recent progress in theory and molecular dynamics simulations as well as in ultrafast vibrational spectroscopy has led to new and detailed insight into fluctuations of water structure, elementary water motions, electric fields at hydrated biointerfaces, and processes of vibrational relaxation and energy dissipation. Here, we review recent advances in both theory and experiment, focusing on hydrated DNA, proteins, and phospholipids, and compare dynamics in the hydration shells to bulk water.

Abstract: Antibody-dependent enhancement (ADE) of viral entry has been a major concern for epidemiology, vaccine development, and antibody-based drug therapy. However, the molecular mechanism behind ADE is still elusive. Coronavirus spike protein mediates viral entry into cells by first binding to a receptor on the host cell surface and then fusing viral and host membranes. In this study, we investigated how a neutralizing monoclonal antibody (MAb), which targets the receptor-binding domain (RBD) of Middle East respiratory syndrome (MERS) coronavirus spike, mediates viral entry using pseudovirus entry and biochemical assays. Our results showed that MAb binds to the virus surface spike, allowing it to undergo conformational changes and become prone to proteolytic activation. Meanwhile, MAb binds to cell surface IgG Fc receptor, guiding viral entry through canonical viral-receptor-dependent pathways. Our data suggest that the antibody/Fc-receptor complex functionally mimics viral receptor in mediating viral entry. Moreover, we characterized MAb dosages in viral-receptor-dependent, Fc-receptor-dependent, and both-receptors-dependent viral entry pathways, delineating guidelines on MAb usages in treating viral infections. Our study reveals a novel molecular mechanism for antibody-enhanced viral entry and can guide future vaccination and antiviral strategies.IMPORTANCE Antibody-dependent enhancement (ADE) of viral entry has been observed for many viruses. It was shown that antibodies target one serotype of viruses but only subneutralize another, leading to ADE of the latter viruses. Here we identify a novel mechanism for ADE: a neutralizing antibody binds to the surface spike protein of coronaviruses like a viral receptor, triggers a conformational change of the spike, and mediates viral entry into IgG Fc receptor-expressing cells through canonical viral-receptor-dependent pathways. We further evaluated how antibody dosages impacted viral entry into cells expressing viral receptor, Fc receptor, or both receptors. This study reveals complex roles of antibodies in viral entry and can guide future vaccine design and antibody-based drug therapy.