搜索结果：250K

Venous thromboembolism (VTE) is a serious complication after hepatopancreatobiliary surgery. We investigated the incidence and timing of VTE events postoperatively and explored factors associated with VTE development. We analyzed the SEER-Medicare database for patients with primary pancreas or liver cancer undergoing resection between 2002 and 2019. The primary outcome was VTE within 90 days postoperatively and the secondary outcomes included 90-day postoperative deep venous thrombosis and pulmonary embolism and late (91-180 days) VTE. We identified 17756 eligible surgeries. The median age was 73 years, 49 % were female, and 59 % underwent a pancreatoduodenectomy. In total, 1695 patients (10 %) developed VTE within six months, with 32 % within one month and 67 % within 3 months postoperatively. Variables independently associated with 90-day VTE included female sex (OR 1.14), higher Elixhauser (OR 1.06) and Caprini scores (OR 1.14), residence in metropolitan area (population >1 million vs < 250k, OR 1.32), dual Medicaid eligibility (OR 3.07), advanced stage (distant vs localized, OR 2.03), surgery type (Whipple vs partial hepatectomy, OR 1.75), and lower hospital volume (1st vs 4th quartile, OR 1.41). Patients with pancreas and liver cancer continued to experience VTE events up to 6 months following surgery. We identified clinical and sociodemographic factors for VTE risk stratification.

Poly(lactic acid) (PLA) has gained substantial attention in the bioplastic industry. However, its practical application is hindered by inherent brittleness and slow biodegradability. This research purposes to simultaneously enhance the toughness and biodegradation rate of PLA through melt blending with 1-5&#xa0;wt% of epoxidized soybean oil (ESO) and epoxidized natural rubbers (ENRs) with molecular weights of &#x223c;70,000&#xa0;g/mol (ENR-70k) and &#x223c;250,000&#xa0;g/mol (ENR-250k). The incorporation of these additives significantly improved the mechanical properties of PLA, increasing elongation by 26-30 times and impact strength by 2.0-3.2 times at 5&#xa0;wt% loading compared to neat PLA. ESO provided the greatest enhancement in stretchability, whereas ENR-250k led to the highest impact strength. These improvements are attributed to enhanced interfacial adhesion between the dispersed phase and PLA matrix, evidenced by downward shifts in glass transition temperature and the formation of yielding and microfibrils on impact-fractured surfaces. Moreover, PLA/ENR-250k blends exhibited maximum biodegradation of &#x223c;74&#xa0;% compared to neat PLA (&#x223c;38&#xa0;%) at 90&#xa0;days, measured from CO2 evolution, which correlates with increased chain mobility and elastomer compatibility. The results highlight the effectiveness of tailored elastomers in promoting the environmental decomposition of PLA, offering a promising pathway for expanding its applicability in sustainable materials design, particularly where accelerated biodegradability is essential.

De novo molecular design remains a fundamental challenge in drug discovery, requiring simultaneous optimization of multiple conflicting objectives such as drug-likeness, synthetic accessibility, and novelty while maintaining chemical validity. We present HybridMolGen, a novel unified framework that synergistically combines three complementary deep learning paradigms: (1) diffusion probabilistic models that generate high-quality, chemically valid molecular samples through gradual noise removal, (2) SE(3)-equivariant graph neural networks that enforce geometric and topological constraints ensuring structural validity and molecular diversity, and (3) property-conditioned transformers that enable fine-grained control over multiple objectives through multi-layer cross-attention modulation. These components operate within a multi-objective reinforcement learning paradigm that discovers optimal property tradeoffs without manual weight tuning. Extensive benchmarking on MOSES, GuacaMol, and ZINC-250k datasets demonstrates state-of-the-art performance: 99.7% validity, 94.3% novelty, average QED score of 0.753, and 4.9% improvement in GuacaMol overall scores. Critically, HybridMolGen discovers 1.57&#xd7; more molecules satisfying all target criteria simultaneously (91.3% versus 58.3% for CPRL) and generates 2.23&#xd7; more Pareto-efficient solutions compared to traditional scalarization, demonstrating genuine architectural synergy beyond simple component aggregation. Comprehensive ablation studies confirm that the three-way integration outperforms even the best two-component combination by 6.5%, positioning HybridMolGen as a powerful tool for accelerating drug discovery pipelines. Implementation code is available as Supplementary Material, available as supplementary data at Bioinformatics online.

Studying replicative aging in yeast is a central component of aging research. Recent advances in time-lapse microscopy and microfluidics now enable continuous, high-resolution tracking of individual yeast cells throughout their lifespan. However, quantifying replicative lifespan from microscopy data remains labor-intensive, as it traditionally requires manual counting of cell division events for each cell. Recent deep learning-based approaches have begun to address this challenge by automating lifespan quantification. Here, we present a versatile image analysis framework that accurately detects yeast cell division events during replicative aging. To reduce the need for large, manually annotated datasets, we pretrain a Masked Autoencoder on large-scale (~250K), unlabeled yeast cell image crops. This self-supervised pretraining substantially lowers the amount of annotated data required to train a transformer model for division event detection. Moreover, our model is trained to directly identify budding events, eliminating dependence on arbitrary heuristics such as changes in cell area. By leveraging self-supervised learning, our approach only requires training data with fewer than 50 mother cells (~1,000 division events, which is significantly lower than reported in previous methods), while maintaining high detection accuracy.

Aquaculture has become a crucial component of global food production, yet catfish (10.8% of global finfish production) breeding programs often lack sufficient genetic data to fully utilize their production potential. In the last 15 years, there have been improvements in this field as two high-density (HD) single nucleotide polymorphism (SNP) arrays (250K and 690K) and low-density panels have been developed for North American channel catfish (Ictalurus punctatus) and blue catfish (Ictalurus furcatus). This lack of genomic tools hinders genetic improvement efforts in other commercially relevant catfish species besides them. Therefore, this review investigated the reason behind the lack of SNP chip usage in genetic-based selections in most catfish breeding programs and the cross-species applicability of the already existing high-density SNP arrays for genotyping members of the Clariidae, African catfish (Clarias gariepinu), and Siluridae, European catfish (Silurus glanis), families. This paper systematically reviews the literature of more than 16 SNP arrays, with 66 non-target species, and assesses the possibility of adapting catfish SNP arrays to the catfish families of interest. With lowered filtering (e.g., MAF > 0) thresholds, the Affymetrix Axiom 250K and Axiom Catfish 690K Genotyping Array could potentially be used on important market species like African and European catfishes. In the long term, chip development would be the solution for these species, but, until then, cross-application is a viable alternative. Despite low polymorphic SNPs (~1%) and call rates (~0%), this SNP array could aid researchers and breeders, improving catfish aquaculture and management.

Atherosclerosis, a major cause of cardiovascular diseases, is characterized by the buildup of lipids and chronic inflammation in the arteries, leading to plaque formation and potential rupture. Despite recent advances in single-cell transcriptomics (scRNA-seq), the underlying immune mechanisms and transformations in structural cells driving plaque progression remain incompletely defined. Existing datasets often lack comprehensive coverage and consistent annotations, limiting the utility of downstream analyses. Here, we present an integrated single-cell atlas of human atherosclerotic plaques, covering roughly 250k high-quality annotated cells. We achieve robust cell type annotations validated by expert consensus and surface protein measurements. Using this atlas, we introduce distinct markers for plaque neutrophils, identify a proangiogenic endothelial cell cluster enriched in advanced lesions, and specialized macrophage subsets. We also establish that fibromyocytes are exclusive to vascular tissue. This comprehensive atlas enables accurate automatic cell type annotation of new datasets, improves experimental design by guiding sample size and detection power, and supports the deconvolution of bulk RNA-seq data. An interactive WebUI makes these resources widely accessible.

Representation learning via pre-trained deep learning models is emerging as an integral method for studying the molecular structure-property relationship, which is then leveraged to predict molecular properties or design new molecules with desired attributes. We propose an unsupervised method to localize and characterize representations of pre-trained models through the lens of non-parametric property-driven subset scanning (PDSS), to improve the interpretability of deep molecular representations. We assess its detection capabilities on diverse molecular benchmarks (ZINC-250K, MOSES, MoleculeNet, FlavorDB, M2OR) across predictive chemical language models (MoLFormer, ChemBERTa) and molecular graph generative models (GraphAF, GCPN). We further study how representations evolve due to domain adaptation, and we evaluate the usefulness of the extracted property-driven elements in the embeddings as lower-dimension representations for downstream tasks. Experiments reveal notable information condensation in the pre-trained embeddings upon task-specific fine-tuning. For example, among the property-driven elements found in the embedding (out of [Formula: see text]), only 11 are shared between three distinct tasks (BACE, BBBP, and HIV), while [Formula: see text]-80 of those are unique to each task. Similar patterns are found for flavor and odor detection tasks. When we use the discovered property-driven elements as features for a new task, we find the same or improved performance (3 points up) while reducing the dimensions by 75% without fine-tuning required, thus indicating information localization.

Cannabis sativa L., known for its medicinal and psychoactive properties, has recently experienced rapid market expansion but remains understudied in terms of its fundamental biology due to historical prohibitions. This pioneering study implements GS and ML to optimize cannabinoid profiles in cannabis breeding. We analyzed a representative population of drug-type cannabis accessions, quantifying major cannabinoids and utilizing high-density genotyping with 250K SNPs for GS. Our evaluations of various models-including ML algorithms, statistical methods, and Bayesian approaches-highlighted Random Forest's superior predictive accuracy for single and multi-trait genomic predictions, particularly for THC, CBD, and their precursors. Multi-trait analyses elucidated complex genetic interdependencies and identified key loci crucial to cannabinoid biosynthesis. These results demonstrate the efficacy of integrating GS and ML in developing cannabis varieties with tailored cannabinoid profiles.

In the digital age, coordinated inauthentic behavior threatens societal stability, markets, and security. Advances in generative AI amplify these threats, enabling effortless content creation, amplifying actors' influence. Detection is hindered by cross-domain activity, where pseudonymous profiles operate across encrypted platforms, and by privacy constraints limiting content analysis. In this study, we propose a robust and scalable cross-domain identity matching framework, based on bursty dynamics, independent of content or interaction data. It outperforms state-of-the-art temporal and structural approaches, remains resilient to incomplete data, and correctly identifies 35% of profiles after 52 weeks. It scales effectively, attaining AUC 0.78 when matching identities across 500 marketplaces with over 250k daily traders. By framing identity matching within the "network of networks" perspective, we demonstrate how coordinated behavior propagates across domains. This dual methodological and theoretical contribution paves the way for innovative strategies to combat digital threats in an increasingly complex and adversarial landscape.

Maximum Power Point Tracking (MPPT) is a technique employed in photovoltaic (PV) systems to ensure that the modules transfer the maximum generated power to the load. An advanced algorithm, the Improved Optimized Adaptive Differential Conductance (IOADC), was developed by applying Kirchhoff's law within a single diode model framework. The algorithm's performance was evaluated under various solar irradiance levels of 500&#xa0;W/m2, 750&#xa0;W/m2, and 1000&#xa0;W/m2 at a constant temperature of 298K, analyzing its impact on power generation and transfer. Additionally, the performance was assessed at varying temperatures of 250K, 298K, and 350K under a constant irradiance of 1000&#xa0;W/m2 to examine its effect on the Module Saturation Current (MSC). The analysis revealed that the PV modules' impedance decreases with increasing irradiance, while the load's impedance remains largely unaffected which aligns with the PV applications. However, the implementation of the IOADC technique showed significant effectiveness. It was also noted that an increase in temperature raises the module saturation current, which in turn reduces the power output, and vice versa which also agrees with the PV application. Real-world application results indicated that at an irradiance of 750&#xa0;W/m2, the output power at the maximum power point (MPP) for the Optimized Adaptive Differential Conductance (OADC), Voltage Control Technique, and IOADC were 83.3346&#xa0;W, 86.9122&#xa0;W, and 100.1739&#xa0;W, respectively. The 100.1739W obtained from the IOADC technique showed a significant improvement. Through comprehensive comparative evaluation, analysis, and validation of the effects of varying temperature, irradiance, and MSC on output power, the developed IOADC model demonstrated a relative improvement of 15.82&#xa0;% in simulations and 20.21&#xa0;% in real-world conditions compared to the Voltage Control Technique and the OADC technique, respectively. Simulation validation and real-world application validation were performed using MATLAB 2020b. These validations confirmed the superior performance of the IOADC algorithm under varying conditions of temperature, irradiance, and module saturation current.

The COVID-19 pandemic disrupted healthcare access and utilization throughout the US, with variable impact on patients of different socioeconomic status (SES) and race. We characterize pre-pandemic and pandemic demographic and SES trends of lumbar fusion patients in the US. Adults undergoing first-time lumbar fusion 1/1/2004-3/31/2021 were assessed in Clinformatics&#xae; Data Mart for patient age, geographical location, gender, race, education level, net worth, and Charlson Comorbidity Index (CCI). Multivariable regression models were used to evaluate the significance of trends over time, with a focus on pandemic trends 2020-2021 versus previous trends 2004-2019. The total 217,204 patients underwent lumbar fusions, 1/1/2004-3/31/2021. The numbers and per capita rates of lumbar fusions increased 2004-2019 and decreased in 2020 (first year of COVID-19 pandemic), with large variation in geographic distribution. There was overall a significant decrease in proportion of White patients undergoing lumbar fusion over time (OR=0.997, p<.001), though they were more likely to undergo surgery during the pandemic (OR=1.016, p<.001). From 2004-2021, patients were more likely to be educated beyond high school. Additionally, patients in the highest (>$500k) and lowest (<$25k) net worth categories had significantly more fusions over time (p<.001). During the pandemic (2020-2021), patients in higher net worth groups were more likely to undergo lumbar fusions ($150k-249k & $250k-499k: p<.001) whereas patients in the lowest net worth group had decreased rate of surgeries (p<.001). Lastly, patients' CCI increased significantly from 2004 to 2021 (coefficient=0.124, p<.001), and this trend held true during the pandemic (coefficient=0.179, p<.001). To the best of our knowledge, our work represents the most comprehensive and recent characterization of SES variables in lumbar fusion rates. Unsurprisingly, lumbar fusions decreased overall with the onset of the COVID-19 pandemic. Importantly, disparities in fusion patients across patient race and wealth widened during the pandemic, reversing years of progress, a lesson we can learn for future public health emergencies.

Deep generative models are becoming a tool of choice for exploring the molecular space. One important application area of deep generative models is the reverse design of drug compounds for given attributes (solubility, ease of synthesis, etc.). Although there are many generative models, these models cannot generate specific intervals of attributes. This paper proposes a AC-ModNet model that effectively combines VAE with AC-GAN to generate molecular structures in specific attribute intervals. The AC-ModNet is trained and evaluated using the open 250K ZINC dataset. In comparison with related models, our method performs best in the FCD and Frag model evaluation indicators. Moreover, we prove the AC-ModNet created molecules have potential application value in drug design by comparing and analyzing them with medical records in the PubChem database. The results of this paper will provide a new method for machine learning drug reverse design.

We present the EuroCity Persons (ECP) 2.0 dataset, a novel image dataset for person detection, tracking and prediction in traffic. The dataset was collected on-board a vehicle driving through 29 cities in 11 European countries. It contains more than 250K unique person trajectories, in more than 2.0M images and comes with a size of 11 TB. ECP2.0 is about one order of magnitude larger than previous state-of-the-art person datasets in automotive context. It offers remarkable diversity in terms of geographical coverage, time of day, weather and seasons. We discuss the novel semi-supervised approach that was used to generate the temporally dense pseudo ground-truth (i.e., 2D bounding boxes, 3D person locations) from sparse, manual annotations at keyframes. Our approach leverages auxiliary LiDAR data for 3D uplifting and vehicle inertial sensing for ego-motion compensation. It incorporates keyframe information in a three-stage approach (tracklet generation, tracklet merging into tracks, track smoothing) for obtaining accurate person trajectories. We validate our pseudo ground-truth generation approach in ablation studies, and show that it significantly outperforms existing methods. Furthermore, we demonstrate its benefits for training and testing of state-of-the-art tracking methods. Our approach provides a speed-up factor of about 34 compared to frame-wise manual annotation. The ECP2.0 dataset is made freely available for non-commercial research use.

Hereditary hearing loss is a genetically heterogeneous neurosensory disorder that affects many people. Deafness and infertility can coexist in some cases, creating the hearing impairment infertile male syndrome. There are several known molecular mechanisms that can cause deafness either on its own or in conjunction with infertility. Here, we represent two consanguineous families (A, B), both families had clinical evidence of deafness, and family B also had infertility, so we referred to them as having nonsyndromic hearing loss (NSHL) and hearing impairment infertile male syndrome (HIIMS), respectively. These families' genetic makeup was examined using an Affymetrix GeneChip 250K Nsp array followed by Sanger sequencing. In family A, we identified a novel homozygous stop gain variant [NM_003672.4; c.1000C>T; p.(Gln334*)] and a homozygous missense variant [NM_003672.4; c.684C>A; p.(Asn228Lys)] in family B in CDC14A gene (MIM#603504). In animal models, the CDC14A gene causes both hearing loss and infertility; in addition, it also causes NSHL and HIIMS in humans. Our study on the CDC14A gene has identified two novel variants, crucial for delineating disease boundaries. Variants in exon 10 and upstream cause HIIMS, and those in exon 11 and downstream are linked exclusively to hearing impairment. This precision enhances diagnostics and offers potential for targeted interventions, marking a significant advancement in understanding the genetic basis of these conditions.

As an appealing approach for discovering novel leads, the key advantage of de novo drug design lies in its ability to explore a much broader dimension of chemical space, without being confined to the knowledge of existing compounds. So far, many generative models have been described in the literature, which have completely redefined the concept of de novo drug design. However, many of them lack practical value for real-world drug discovery. In this work, we have developed a graph-based generative model within a reinforcement learning framework, namely, METEOR (Molecular Exploration Through multiplE-Objective Reinforcement). The backend agent of METEOR is based on the well-established GCPN model. To ensure the overall quality of the generated molecular graphs, we implemented a set of rules to identify and exclude undesired substructures. Importantly, METEOR is designed to conduct multi-objective optimization, i.e., simultaneously optimizing binding affinity, drug-likeness, and synthetic accessibility of the generated molecules under the guidance of a special reward function. We demonstrate in a specific test case that without prior knowledge of true binders to the chosen target protein, METEOR generated molecules with superior properties compared to those in the ZINC 250k data set. In conclusion, we have demonstrated the potential of METEOR as a practical tool for generating rational drug-like molecules in the early phase of drug discovery.

The temperature-dependent kinetic parameters, branching fractions, and chaperone effects of the self- and cross-reactions between acetonyl peroxy (CH3C(O)CH2O2) and hydro peroxy (HO2) have been studied using pulsed laser photolysis coupled with infrared (IR) wavelength-modulation spectroscopy and ultraviolet absorption (UVA) spectroscopy. Two IR lasers simultaneously monitored HO2 and hydroxyl (OH), while UVA measurements monitored CH3C(O)CH2O2. For the CH3C(O)CH2O2 self-reaction (T = 270-330 K), the rate parameters were determined to be A = (1.5-0.3+0.4) &#xd7; 10-13 and Ea/R = -996 &#xb1; 334 K and the branching fraction to the alkoxy channel, k2b/k2, showed an inverse temperature dependence following the expression, k2b/k2 = (2.27 &#xb1; 0.62) - [(6.35 &#xb1; 2.06) &#xd7; 10-3] T(K). For the reaction between CH3C(O)CH2O2 and HO2 (T = 270-330 K), the rate parameters were determined to be A = (3.4-1.5+2.5) &#xd7; 10-13 and Ea/R = -547 &#xb1; 415 K for the hydroperoxide product channel and A = (6.23-4.4+15.3) &#xd7; 10-17 and Ea/R = -3100 &#xb1; 870 K for the OH product channel. The branching fraction for the OH channel, k1b /k1, follows the temperature-dependent expression, k1b/k1 = (3.27 &#xb1; 0.51) - [(9.6 &#xb1; 1.7) &#xd7; 10-3] T(K). Determination of these parameters required an extensive reaction kinetics model which included a re-evaluation of the temperature dependence of the HO2 self-reaction chaperone enhancement parameters due to the methanol-hydroperoxy complex. The second-law thermodynamic parameters for KP,M for the formation of the complex were found to be &#x394;rH250K&#xb0; = -38.6 &#xb1; 3.3 kJ mol-1 and &#x394;rS250K&#xb0; = -110.5 &#xb1; 13.2 J mol-1 K-1, with the third-law analysis yielding &#x394;rH250K&#xb0; = -37.5 &#xb1; 0.25 kJ mol-1. The HO2 self-reaction rate coefficient was determined to be k4 = (3.34-0.80+1.04) &#xd7; 10-13 exp [(507 &#xb1; 76)/T]cm3 molecule-1 s-1 with the enhancement term k4,M&#x2033; = (2.7-1.7+4.7) &#xd7; 10-36 exp [(4700 &#xb1; 255)/T]cm6 molecule-2 s-1, proportional to [CH3OH], over T = 220-280 K. The equivalent chaperone enhancement parameter for the acetone-hydroperoxy complex was also required and determined to be k4,A&#x2033; = (5.0 &#xd7; 10-38 - 1.4 &#xd7; 10-41) exp[(7396 &#xb1; 1172)/T] cm6 molecule-2 s-1, proportional to [CH3C(O)CH3], over T = 270-296 K. From these parameters, the rate coefficients for the reactions between HO2 and the respective complexes over the given temperature ranges can be estimated: for HO2&#xb7;CH3OH, k12 = [(1.72 &#xb1; 0.050) &#xd7; 10-11] exp [(314 &#xb1; 7.2)/T] cm3 molecule-1 s-1 and for HO2&#xb7;CH3C(O)CH3, k15 = [(7.9 &#xb1; 0.72) &#xd7; 10-17] exp [(3881 &#xb1; 25)/T] cm3 molecule-1 s-1. Lastly, an estimate of the rate coefficient for the HO2&#xb7;CH3OH self-reaction was also determined to be k13 = (1.3 &#xb1; 0.45) &#xd7; 10-10 cm3 molecule-1 s-1.

To determine the association between academic productivity and industry compensation among Orthopaedic Traumatologists. Retrospective cohort study. Review of the Centers for Medicaid and Medicare Services Open Payments program from 2016 to 2020. 1120 Orthopaedic Traumatologists. To determine if an Orthopaedic Traumatologist's h-index and m-index, as generated from Web of Science, Scopus, and Google Scholar User Profile databases, correlate with total payments from medical industry in 7 categories, including Royalties and Licensing Fees, Consulting Fees, Gifts, Honoraria, and 3 unique Speaking Fee delineations. Of 30,343 Orthopaedic Surgeons in the Open Payments program, 1120 self-identified with the Orthopaedic Trauma taxonomy. From 2016 to 2020, 499 surgeons (44.6%) received compensation in one of the eligible categories, most commonly from Consulting Fees (67.3%), though payments from Royalties provided the greatest gross income (70.4%). Overall, for all 1120 surgeons, h-index (r = 0.253, P < 0.001) and m-index (r = 0.136, P < 0.01) correlated positively with mean annual total industry compensation. The highest annual compensation group had higher h-index ($0 vs. $1-$1k vs. $1k-$10k vs. >$10k: 5.0 vs. 6.6 vs. 9.6 vs. 16.8, P < 0.001) and m-index ($0 vs. $1-$1k vs. $1k-$10k vs. >$10k: 0.48 vs. 0.60 vs. 0.65 vs. 0.89, P < 0.001) scores than either the intermediate or the no compensation groups. Multivariable analysis of factors associated with increased industry compensation, including H-index and years active, identified both as having significant associations with physician payments [H-index (B = 0.073, P < 0.001); years active (B = 0.059, P < 0.001)]. Subgroup analysis of the highest annual earner group (>$250k/year) also demonstrated the highest overall h-index (27.6, P < 0.001) and m-index (1.23, P = 0.047) scores, even when compared with other high-earners ($10k-$50k, $50k-$250k). Overall, each increase in h-index above an h-index of 3 was associated with an additional $1722 (95% CI: $1298-2146) of annual industry compensation. Academic productivity metrics have a positive association with industry compensation for Orthopaedic Traumatologists. This may highlight a potential ancillary benefit to scholarly efforts.

There are limited data regarding the impact of gender within congenital heart surgery. Our aim was to assess gender-related experiences by surgeons in this field. A cross-sectional survey was emailed to practicing congenital heart surgeons to ascertain the perception of gender in 5 domains: training, professional career, clinical practice, personal life, and career outlook. The survey response rate was 94% (17/18) for women and 44% (112/257) for men. More than half of women (53%) were discouraged from pursuing congenital heart surgery (P&#xa0;<&#xa0;.001) and reported a negative impact of gender in attaining their first congenital heart surgery job (P&#xa0;<&#xa0;.001) compared with men. Despite similar demographics, women reported lower starting annual salaries ($150K-$250K vs $250K-$400K), lower current annual salaries ($500K-$750K vs $750K-$1M), lower academic ranks (clinical instructor 6% vs 4% [P&#xa0;=&#xa0;.045], assistant professor 35% vs 19% [P&#xa0;=&#xa0;.19], associate professor 41% vs 25% [P&#xa0;=&#xa0;.24], and professor 6% vs 41% [P&#xa0;=&#xa0;.005]) along with lower annual salaries at the associate professor ($500K-$750K vs $1M-$1.25M) and professor levels ($1M-$1.25M vs >$1.5M) compared with men. Sexual harassment was experienced more frequently by women both in training (65% vs 6%, P&#xa0;<&#xa0;.001) and in practice (65% and 4%, P&#xa0;<&#xa0;.001). This survey highlights many areas of gender-related differences: discouragement due to gender to pursue congenital heart surgery, sexual harassment in training and practice, salary and academic rank differentials, negative gender perception at work, and lower career satisfaction for women. Despite various differences between both genders, the majority in each group would choose to enter this profession again as well as encourage others to do so.

Generating new molecules with the desired physical or chemical properties is the key challenge of computational material design. Deep learning techniques are being actively applied in the field of data-driven material informatics and provide a promising way to accelerate the discovery of innovative materials. In this work, we utilize an invertible graph generative model to generate hypothetical promising high-temperature polymer dielectrics. A molecular graph generative model based on the invertible normalizing flow is trained on a data set containing 250k polymer molecular graphs (mostly generated by an RNN-based generative model) to learn the invertible transformations between latent distributions and molecular graph structures. When generating molecular graphs, a sample vector is drawn from the latent space, and then an adjacency tensor and node attribute matrix are generated through two invertible flows in two steps and assembled into a molecular graph. The model has the merits of exact likelihood training and an efficient one-shot generation process. The learned latent space is used to generate polymers with a high glass-transition temperature (Tg) and a wide band gap (Eg) for the application of high-temperature energy storage film capacitors. This work contributes to the efficient design of high-temperature polymer dielectrics by using deep generative models.