Search

Deep learning techniques have recently demonstrated remarkable success in numerous domains. Typically, the success of these deep learning models is measured in terms of performance metrics such as accuracy and mean average precision (mAP). Generally, a model’s high performance is highly valued, but it frequently comes at the expense of substantial energy costs and carbon footprint emissions during the model building step. Massive emission of CO2 has a deleterious impact on life on earth in general and is a serious ethical concern that is largely ignored in deep learning research. In this article, we mainly focus on environmental costs and the means of mitigating carbon footprints in deep learning models, with a particular focus on models created using knowledge distillation (KD). Deep learning models typically contain a large number of parameters, resulting in a ‘heavy’ model. A heavy model scores high on performance metrics but is incompatible with mobile and edge computing devices. Model compression techniques such as knowledge distillation enable the creation of lightweight, deployable models for these low-resource devices. KD generates lighter models and typically performs with slightly less accuracy than the heavier teacher model (model accuracy by the teacher model on CIFAR 10, CIFAR 100, and TinyImageNet is 95.04%, 76.03%, and 63.39%; model accuracy by KD is 91.78%, 69.7%, and 60.49%). Although the distillation process makes models deployable on low-resource devices, they were found to consume an exorbitant amount of energy and have a substantial carbon footprint (15.8, 17.9, and 13.5 times more carbon compared to the corresponding teacher model). The enormous environmental cost is primarily attributable to the tuning of the hyperparameter, Temperature (τ). In this article, we propose measuring the environmental costs of deep learning work (in terms of GFLOPS in millions, energy consumption in kWh, and CO2 equivalent in grams). In order to create lightweight models with low environmental costs, we propose a straightforward yet effective method for selecting a hyperparameter (τ) using a stochastic approach for each training batch fed into the models. We applied knowledge distillation (including its data-free variant) to problems involving image classification and object detection. To evaluate the robustness of our method, we ran experiments on various datasets (CIFAR 10, CIFAR 100, Tiny ImageNet, and PASCAL VOC) and models (ResNet18, MobileNetV2, Wrn-40-2). Our novel approach reduces the environmental costs by a large margin by eliminating the requirement of expensive hyperparameter tuning without sacrificing performance. Empirical results on the CIFAR 10 dataset show that the stochastic technique achieves an accuracy of 91.67%, whereas tuning achieves an accuracy of 91.78%—however, the stochastic approach reduces the energy consumption and CO2 equivalent each by a factor of 19. Similar results have been obtained with CIFAR 100 and TinyImageNet dataset. This pattern is also observed in object detection classification on the PASCAL VOC dataset, where the tuning technique performs similarly to the stochastic technique, with a difference of 0.03% mAP favoring the stochastic technique while reducing the energy consumptions and CO2 emission each by a factor of 18.5.

The Percepta Genomic Sequencing Classifier (GSC) was developed to up-classify as well as down-classify the risk of malignancy for lung lesions when bronchoscopy is non-diagnostic. We evaluated the performance of Percepta GSC in risk re-classification of indeterminate lung lesions. This multicenter study included individuals who currently or formerly smoked undergoing bronchoscopy for suspected lung cancer from the AEGIS I/ II cohorts and the Percepta Registry. The classifier was measured in normal-appearing bronchial epithelium from bronchial brushings. The sensitivity, specificity, and predictive values were calculated using predefined thresholds. The ability of the classifier to decrease unnecessary invasive procedures was estimated. A set of 412 patients were included in the validation (prevalence of malignancy was 39.6%). Overall, 29% of intermediate-risk lung lesions were down-classified to low-risk with a 91.0% negative predictive value (NPV) and 12.2% of intermediate-risk lesions were up-classified to high-risk with a 65.4% positive predictive value (PPV). In addition, 54.5% of low-risk lesions were down-classified to very low risk with99% NPV and 27.3% of high-risk lesions were up-classified to very high risk with a 91.5% PPV. If the classifier results were used in nodule management, 50% of patients with benign lesions and 29% of patients with malignant lesions undergoing additional invasive procedures could have avoided these procedures. The Percepta GSC is highly accurate as both a rule-out and rule-in test. This high accuracy of risk re-classification may lead to improved management of lung lesions.

Objectives Burden of disease estimation commonly requires estimates of the population exposed to a risk factor over a time window (yeart to yeart+n). We present a microsimulation modelling approach for producing such estimates and apply it to calculate the population exposed to long working hours for one country (Italy). Methods We developed a three-model approach: Model 1, a multilevel model, estimates exposure to the risk factor at the first year of the time window (yeart). Model 2, a regression model, estimates transition probabilities between exposure categories during the time window (yeart to yeart+n). Model 3, a microsimulation model, estimates the exposed population over the time window, using the Monte Carlo method. The microsimulation is carried out in three steps: (a) a representative synthetic population is initiated in the first year of the time window using prevalence estimates from Model 1, (b) the exposed population is simulated over the time window using the transition probabilities from Model 2; and (c) the population is censored for deaths during the time window. Results We estimated the population exposed to long working hours (i.e. 41–48, 49–54 and ≥55 hours/week) over a 10-year time window (2002–11) in Italy. We populated all three models with official data from Labour Force Surveys, United Nations population estimates and World Health Organization life tables. Estimates were produced of populations exposed over the time window, disaggregated by sex and 5-year age group. Conclusions Our modelling approach for estimating the population exposed to a risk factor over a time window is simple, versatile, and flexible. It however requires longitudinal exposure data and Model 3 (the microsimulation model) is stochastic. The approach can improve accuracy and transparency in exposure and burden of disease estimations. To improve the approach, a logical next step is changing Model 3 to a deterministic microsimulation method, such as modelling of microflows.

In Morocco, the prevalence of multidrug resistant tuberculosis (MDR-TB) continues to increase especially within previously treated cases; these MDR cases may evolve to extensively drug resistant tuberculosis (XDR-TB) raising major concern to TB control programs. From an epidemiological window, scarce informations are available about the genetic diversity of Mycobacterium tuberculosis (MTB) strains fueling these forms of resistance. The aim of this study was to assess to genetic diversity of MDR-MTB strains. Hence, this prospective study was conducted on patients diagnosed with MDR-TB at Pasteur Institute of Casablanca from 2010 to 2013. A total of 70 MDR-MTB isolates were genotyped by spoligotyping and 15-loci MIRU-VNTR methods. Spoligotyping generated four orphan patterns, five unique profiles whereas 61 strains were grouped in nine clusters (2 to 25 strains per cluster), the clustering rates being 87.1%. Subtyping by 15 loci MIRU-VNTR splitted all clusters already established by spoligotyping and generated 70 unique profiles not recognized in SITVIT2 database; clustering rate was equal to zero. HGDI analysis of 15 loci MIRU demonstrated that eight out of 15 loci were highly discriminant. Of note, all pre-XDR strains belongs to many clades, meaning that there no association between gyrA mutants and particular clade. Overall, the data generated by this study (i) describe the population structure of MDR MTBC in Morocco which is highly homogenous, (ii) confirm that TB in Morocco is almost exclusively transmitted by modern and evolutionary lineages with high level of biodiversity seen by MIRU, and (iii) validate the use of optimized 15-loci MIRU-VNTR format for future investigations in Morocco.

Random regression models (RRM) are used extensively for genomic inference and prediction of time-valued traits in animal breeding, but only recently have been used in plant systems. High-throughput phenotyping (HTP) platforms provide a powerful means to collect high-dimensional phenotypes throughout the growing season for large populations. However, to date, selection of an appropriate statistical genomic framework to integrate multiple temporal traits for genomic prediction in plants remains unexplored. Here, we demonstrate the utility of a multi-trait RRM (MT-RRM) for genomic prediction of daily water usage (WU) in rice (Oryza sativa) through joint modeling with shoot biomass (projected shoot area, PSA). Three hundred and fifty-seven accessions were phenotyped daily for WU and PSA over 20 days using a greenhouse-based HTP platform. MT-RRMs that modeled additive genetic and permanent environmental effects for both traits using quadratic Legendre polynomials were used to assess genomic correlations between traits and genomic prediction for WU. Predictive abilities of the MT-RRMs were assessed using two cross-validation (CV) scenarios. The first scenario was designed to predict genetic values for WU at all time points for a set of accessions with unobserved WU. The second scenario was designed to forecast future genetic values for WU for a panel of known accessions with records for WU at earlier time periods. In each scenario we evaluated two MT-RRMs in which PSA records were absent or available for time points in the testing population. Moderate to strong genomic correlations between WU and PSA were observed across the days of imaging (0.29-0.87). In both CV scenarios, MT-RRMs showed better predictive abilities compared to single-trait RRM, and prediction accuracies were greatly improved when PSA records were available for the testing population. In summary, these frameworks provide an effective approach to predict temporal physiological traits that are difficult or expensive to quantify in large populations.

Protein L-isoaspartyl methyltransferase (PIMT/PCMT1), a product of the pcmt1 gene, catalyzes repair of abnormal L-isoaspartyl linkages in age-damaged proteins. Pcmt1 knockout mice exhibit a profound neuropathology and die 30-60 days postnatal from an epileptic seizure. Here we characterize four new SNP variants of human PIMT with respect to enzymatic activity, thermal stability, and propensity to aggregation. Under standard assay conditions, L191S, A150V, P174H and A65V showed activity losses of 72%, 64%, 61%, and 11% respectively. By differential scanning fluorimetry, melting temperature deviations were -5.2, -4.5, +0.5, and -3.4°C. SDS-PAGE of purified protein reveal significant aggregation of L191S, A150V, and P174H, but not A65V. We also report new data on three unusual PIMT variants among the 13 recently characterized by our laboratory. A7P and I58V were previously found to have 1.8-2.0 times the activity of WT PIMT in the standard assay; however, upon kinetic analysis, we find both variants exhibit reduced catalytic efficiency (Vmax/Km) due to weak isoaspartyl substrate binding. The near complete loss of activity (

Objective To understand the molecular pathways underlying the cardiac preconditioning effect of short-term caloric restriction (CR). Background Lifelong CR has been suggested to reduce the incidence of cardiovascular disease through a variety of mechanisms. However, prolonged adherence to a CR life-style is difficult. Here we reveal the pathways that are modulated by short-term CR, which are associated with protection of the mouse heart from ischemia. Methods Male 10-12 wk old C57bl/6 mice were randomly assigned to an ad libitum (AL) diet with free access to regular chow, or CR, receiving 30% less food for 7 days (d), prior to myocardial infarction (MI) via permanent coronary ligation. At d8, the left ventricles (LV) of AL and CR mice were collected for Western blot, mRNA and microRNA (miR) analyses to identify cardioprotective gene expression signatures. In separate groups, infarct size, cardiac hemodynamics and protein abundance of caspase 3 was measured at d2 post-MI. Results This short-term model of CR was associated with cardio-protection, as evidenced by decreased infarct size (18.5±2.4% vs. 26.6±1.7%, N=10/group; P=0.01). mRNA and miR profiles pre-MI (N=5/group) identified genes modulated by short-term CR to be associated with circadian clock, oxidative stress, immune function, apoptosis, metabolism, angiogenesis, cytoskeleton and extracellular matrix (ECM). Western blots pre-MI revealed CR-associated increases in phosphorylated Akt and GSK3ß, reduced levels of phosphorylated AMPK and mitochondrial related proteins PGC-1α, cytochrome C and cyclooxygenase (COX) IV, with no differences in the levels of phosphorylated eNOS or MAPK (ERK1/2; p38). CR regimen was also associated with reduced protein abundance of cleaved caspase 3 in the infarcted heart and improved cardiac function.

Funding for Open Access provided by the UMD Libraries Open Access Publishing Fund., Decline of sugar maple in North American forests has been attributed to changes in soil calcium (Ca) and nitrogen (N) by acidic precipitation. Although N is an essential and usually a limiting factor in forests, atmospheric N deposition may cause N-saturation leading to loss of soil Ca. Such changes can affect carbon gain and growth of sugar maple trees and seedlings. We applied a 2^2 factorial arrangement of N and dolomitic limestone containing Ca and Magnesium (Mg) to 12 forest plots in the Catskill Mountain region of NY, USA. To quantify the short-term effects, we measured photosynthetic-light responses of sugar maple mature trees and seedlings two or three times during two summers. We estimated maximum net photosynthesis (An-max) and its related light intensity (PAR at An-max), apparent quantum efficiency (Aqe), and light compensation point (LCP). To quantify the long-term effects, we measured basal area of living mature trees before and 4 and 8 years after treatment applications. Soil and foliar chemistry variables were also measured. Dolomitic limestone increased Ca, Mg, and pH in the soil Oe horizon. Mg was increased in the B horizon when comparing the plots receiving N with those receiving CaMg. In mature trees, foliar Ca and Mg concentrations were higher in the CaMg and N+CaMg plots than in the reference or N plots; foliar Ca concentration was higher in the N+CaMg plots compared with the CaMg plots, foliar Mg was higher in the CaMg plots than the N+CaMg plots; An-max was maximized due to N+CaMg treatment; Aqe decreased by N addition; and PAR at An-max increased by N or CaMg treatments alone, but the increase was maximized by their combination. No treatment effect was detected on basal areas of living mature trees four or eight years after treatment applications. In seedlings, An-max was increased by N+CaMg addition. The reference plots had an open herbaceous layer, but the plots receiving N had a dense monoculture of common woodfern in the forest floor, which can impede seedling survival.

Several studies on pregnancy-associated cancers have suggested an association with congenital anomalies in offspring. Previous studies have included maternal cancers diagnosed up to 2 years after pregnancy; however, long latency periods of some cancers mean that cancers diagnosed many years postpartum might have been present during pregnancy in a preclinical state. This paper considers the association between maternal cancers diagnosed from 2 years prior to pregnancy until the mother reaches 50 years of age, and congenital anomalies, as diagnosed at birth or within the first year of life. The current population-based study looks at associations of cancers in mothers with congenital anomalies in their children. Children were followed up from birth to diagnosis of a congenital anomaly, death, emigration or end of follow-up (whichever occurred first). A total of 56,016 children (2.6%) were considered exposed to a maternal cancer of any type; and they had a hazard ratio (HR) of 1.04 (95% confidence interval [CI]: 1.00, 1.09) compared with unexposed children. The greatest HR was seen among children whose mothers had been diagnosed with cancers before or during pregnancy (HR: 1.37, 95% CI: 1.07, 1.75). Similar results were seen when paternal cancers were used as a 'negative control'. Statistically significant associations were seen for some specific congenital anomalies of organ systems (congenital anomalies of the musculoskeletal system [HR: 1.13, 95% CI: 1.02, 1.25]) and for some specific types of maternal cancer (leukaemia [HR: 1.31, 95% CI: 1.01, 1.61], The results of the main analyses suggest a small increase in risk of congenital anomalies in offspring of mothers diagnosed with cancer from 2 years before pregnancy, until the mother reaches 50 years of age; with the greatest increase seen for exposure in the pre-pregnancy and pregnancy period. These results may reflect shared causes for some cancers and some congenital anomalies. The similar results seen for paternal cancers indicate that the cause may be genetic or related to the families' social and environmental conditions.

BackgroundsWith 10% of the general population aged 15-59 years chronically infected with hepatitis C virus (HCV), Egypt is the country with the highest HCV prevalence worldwide. Healthcare workers (HCWs) are therefore at particularly high risk of HCV infection. Our aim was to study HCV infection risk after occupational blood exposure among HCWs in Cairo.Methodology/principal findingsThe study was conducted in 2008-2010 at Ain Shams University Hospital, Cairo. HCWs reporting an occupational blood exposure at screening, having neither anti-HCV antibodies (anti-HCV) nor HCV RNA, and exposed to a HCV RNA positive patient, were enrolled in a 6-month prospective cohort with follow-up visits at weeks 2, 4, 8, 12 and 24. During follow-up, anti-HCV, HCV RNA and ALT were tested. Among 597 HCWs who reported a blood exposure, anti-HCV prevalence at screening was 7.2%, not different from that of the general population of Cairo after age-standardization (11.6% and 10.4% respectively, p = 0.62). The proportion of HCV viremia among index patients was 37%. Of 73 HCWs exposed to HCV RNA from index patients, nine (12.3%; 95%CI, 5.8-22.1%) presented transient viremia, the majority of which occurred within the first two weeks after exposure. None of the workers presented seroconversion or elevation of ALT.Conclusions/significanceHCWs of a general University hospital in Cairo were exposed to a highly viremic patient population. They experienced frequent occupational blood exposures, particularly in early stages of training. These exposures resulted in transient viremic episodes without established infection. These findings call for further investigation of potential immune protection against HCV persistence in this high risk group.

Rationale Integrin-linked kinase (ILK) has been proposed as a novel molecular target that has translational potential in diverse cardiac diseases, since its upregulation promotes a broadly cardioprotective phenotype. However, ILK has been implicated as both a cardioprotective and oncogenic target, which imposes therapeutic constraints that are generally relevant to the translational potential of many kinases. Objective To study the cardioprotective properties of the activation-resistant, non-oncogenic, mutation of ILK (ILK(R211A)) against experimental MI in vivo and Doxorubicin induced apoptosis in vitro and it's relationships to stress induced heat shock proteins. Methods/results The transgenic mouse heart over-expressing a point mutation in the ILK pleckstrin homology (PH) domain (Tg(R211A)) exhibits a highly cardioprotective phenotype based on LAD-ligation-induced MI reduction in vivo, and on protection against doxorubicin (DOX)-induced cardiomyocyte apoptosis when overexpressed in human induced pluripotent stem cell (iPS)-derived cardiomyocytes in vitro. Intriguingly, the degree of cardioprotection seen with the ILK(R211A) mutation exceeded that with the ILK(S343D) mutation. Microarray and immunoprecipitation analyses revealed upregulation of expression levels and specific binding of ILK(WT), ILK(S343D) and ILK(R211A) to both constitutively active heat-shock protein 70 (Hsc70) and inducible Hsp70 in response to MI, and to acute ILK overexpression in iPSC-cardiomyocytes. ILK-mediated cardioprotection was shown to depend upon Hsp70 ATPase activity. Conclusions These findings indicate that wild type ILK and the non-oncogenic ILK(R211A) mutation comprise a cardioprotective module with Hsp/c70. These results advance a novel target discovery theme in which kinase mutations can be safely engineered to enhance cardioprotective effects.

Even though the total SA American population is increasing rapidly, there is a paucity of information on the relationship between diet quality, acculturation and health outcomes such as Metabolic Syndrome (MetS) in the low-income South Asian (SA) sub-population. Our goal was to examine diet quality, degree of acculturation and their potential influence on MetS in a diverse sample of SA Americans. A convenience sample of 401 adult SA men and women were studied using a cross-sectional study design. Volunteers from two low-income community health clinics in Maryland were interviewed by questionnaires. MetS, defined by the consensus harmonized definition by the presence of ≥ 3 of the 5 abnormal indicators, was studied. An interviewer obtained an automated self-administered 24-hour Recall (ASA24) and an acculturation index (using a previously validated (SL-ASIA). SA had a composite HEI2010 score of 68 suggesting an overall need for diet improvements. Males had a higher diet quality (mean HEI2010 score) than females. Males with MetS had lower diet quality (68) than males without MetS (73). The converse was true for females (68 vs. 65). Americanized (more acculturated) subjects had a higher diet quality compared to less acculturated SA. Small differences were found in diet quality scores among SA adults from different countries. Less acculturated females, had a higher percentage of MetS and lower diet quality compared to males. These results suggest that interventions are needed in males and females who were less acculturated because they may have greater MetS and lower diet quality compared to more Americanized SA.

Background Attrition of students from aviation training is a serious financial and operational concern for the U.S. Navy. Each late stage navy aviator training failure costs the taxpayer over $1,000,000 and ultimately results in decreased operational readiness of the fleet. Currently, potential aviators are selected based on the Aviation Selection Test Battery (ASTB), which is a series of multiple-choice tests that evaluate basic and aviation-related knowledge and ability. However, the ASTB does not evaluate a person's response to stress. This is important because operating sophisticated aircraft demands exceptional performance and causes high psychological stress. Some people are more resistant to this type of stress, and consequently better able to cope with the demands of naval aviation, than others. Methodology/Principal Findings Although many psychological studies have examined psychological stress resistance none have taken advantage of the human genome sequence. Here we use high-throughput -omic biology methods and a novel statistical data normalization method to identify plasma proteins associated with human performance under psychological stress. We identified proteins involved in four basic physiological processes: innate immunity, cardiac function, coagulation and plasma lipid physiology. Conclusions/Significance The proteins identified here further elucidate the physiological response to psychological stress and suggest a hypothesis that stress-susceptible pilots may be more prone to shock. This work also provides potential biomarkers for screening humans for capability of superior performance under stress.