Explore the vast range of titles available.

For its robust predictive power (compared to pure physics-based models) and sample-efficient training (compared to pure deep learning models), physics-informed deep learning (PIDL), a paradigm hybridizing physics-based models and deep neural networks (DNNs), has been booming in science and engineering fields. One key challenge of applying PIDL to various domains and problems lies in the design of a computational graph that integrates physics and DNNs. In other words, how the physics is encoded into DNNs and how the physics and data components are represented. In this paper, we offer an overview of a variety of architecture designs of PIDL computational graphs and how these structures are customized to traffic state estimation (TSE), a central problem in transportation engineering. When observation data, problem type, and goal vary, we demonstrate potential architectures of PIDL computational graphs and compare these variants using the same real-world dataset.

Background Interest in single-cell transcriptomic analysis is growing rapidly, especially for profiling rare or heterogeneous populations of cells. In almost all reported works investigators have used live cells, which introduces cell stress during preparation and hinders complex study designs. Recent studies have indicated that cells fixed by denaturing fixative can be used in single-cell sequencing, however they did not usually work with most types of primary cells including immune cells. Methods The methanol-fixation and new processing method was introduced to preserve human peripheral blood mononuclear cells (PBMCs) for single-cell RNA sequencing (scRNA-Seq) analysis on 10× Chromium platform. Results When methanol fixation protocol was broken up into three steps: fixation, storage and rehydration, we found that PBMC RNA was degraded during rehydration with PBS, not at cell fixation and up to 3-month storage steps. Resuspension but not rehydration in 3× saline sodium citrate (SSC) buffer instead of PBS preserved PBMC RNA integrity and prevented RNA leakage. Diluted SSC buffer did not interfere with full-length cDNA synthesis. The methanol-fixed PBMCs resuspended in 3× SSC were successfully implemented into 10× Chromium standard scRNA-seq workflows with no elevated low quality cells and cell doublets. The fixation process did not alter the single-cell transcriptional profiles and gene expression levels. Major subpopulations classified by marker genes could be identified in fixed PBMCs at a similar proportion as in live PBMCs. This new fixation processing protocol also worked in several other fixed primary cell types and cell lines as in live ones. Conclusions We expect that the methanol-based cell fixation procedure presented here will allow better and more effective batching schemes for a complex single cell experimental design with primary cells or tissues.

The Waymo Open Dataset has been released recently, providing a platform to crowdsource some fundamental challenges for automated vehicles (AVs), such as 3D detection and tracking. While the dataset provides a large amount of high-quality and multi-source driving information, people in academia are more interested in the underlying driving policy programmed in Waymo self-driving cars, which is inaccessible due to AV manufacturers’ proprietary protection. Accordingly, academic researchers have to make various assumptions to implement AV components in their models or simulations, which may not represent the realistic interactions in real-world traffic. Thus, this paper introduces an approach to learn a long short-term memory (LSTM)-based model for imitating the behavior of Waymo’s self-driving model. The proposed model has been evaluated based on Mean Absolute Error (MAE). The experimental results show that our model outperforms several baseline models in driving action prediction. In addition, a visualization tool is presented for verifying the performance of the model.

Corticosteroids have been used for decades to modulate inflammation therapeutically, yet there is a paucity of data on their effects in humans. We examined the changes in cellular and molecular immune system parameters, or “immunome”, in healthy humans after systemic corticosteroid administration. We used multiplexed techniques to query the immunome in 20 volunteers at baseline, and after intravenous hydrocortisone (HC) administered at moderate (250 mg) and low (50 mg) doses, to provide insight into how corticosteroids exert their effects. We performed comprehensive phenotyping of 120 lymphocyte subsets by high dimensional flow cytometry, and observed a decline in circulating specific B and T cell subsets, which reached their nadir 4–8 hours after administration of HC. However, B and T cells rebounded above baseline 24 hours after HC infusion, while NK cell numbers remained stable. Whole transcriptome profiling revealed down regulation of NF-κB signaling, apoptosis, and cell death signaling transcripts that preceded lymphocyte population changes, with activation of NK cell and glucocorticoid receptor signaling transcripts. Our study is the first to systematically characterize the effects of corticosteroids on the human immunome, and we demonstrate that HC exerts differential effects on B and T lymphocytes and natural killer cells in humans.

Background Clinical CAR T-cell therapy using integrating vector systems represents a promising approach for the treatment of hematological malignancies. Lentiviral and γ-retroviral vectors are the most commonly used vectors in the manufacturing process. However, the integration pattern of these viral vectors and subsequent effect on CAR T-cell products is still unclear. Methods We used a modified viral integration sites analysis (VISA) pipeline to evaluate viral integration events around the whole genome in pre-infusion CAR T-cell products. We compared the differences of integration pattern between lentiviral and γ-retroviral products. We also explored whether the integration sites correlated with clinical outcomes. Results We found that γ-retroviral vectors were more likely to insert than lentiviral vectors into promoter, untranslated, and exon regions, while lentiviral vector integration sites were more likely to occur in intron and intergenic regions. Some integration events affected gene expression at the transcriptional and post-transcriptional level. Moreover, γ-retroviral vectors showed a stronger impact on the host transcriptome. Analysis of individuals with different clinical outcomes revealed genes with differential enrichment of integration events. These genes may affect biological functions by interrupting amino acid sequences and generating abnormal proteins, instead of by affecting mRNA expression. These results suggest that vector integration is associated with CAR T-cell efficacy and clinical responses. Conclusion We found differences in integration patterns, insertion hotspots and effects on gene expression vary between lentiviral and γ-retroviral vectors used in CAR T-cell products and established a foundation upon which we can conduct further analyses.

ObjectiveRecent clinical trials have demonstrated that colchicine may have metabolic and cardiovascular and benefits in at-risk patients; however, the mechanisms through which colchicine may improve outcomes are still unclear. We sought to examine colchicine’s effects on circulating inflammatory and metabolic molecules in adults with obesity and metabolic syndrome (MetS).MethodsBlood samples were collected pre- and post-intervention during a double-blind randomized controlled trial in which 40 adults with obesity and MetS were randomized to colchicine 0.6 mg or placebo twice-daily for 3 months. Serum samples were analyzed for 1305 circulating factors using the SomaScan Platform. The Benjamini–Hochberg procedure was used to adjust the false discovery rate (FDR) for multiple testing.ResultsAt baseline, age (48.0 ± 13.8 vs. 44.7 ± 10.3 years) and BMI (39.8 ± 6.4 vs. 41.8 ± 8.2 kg/m2) were not different between groups. After controlling for the FDR, 34 molecules were significantly changed by colchicine. Colchicine decreased concentrations of multiple inflammatory molecules, including C-reactive protein, interleukin 6, and resistin, in addition to vascular-related proteins (e.g., oxidized low-density lipoprotein receptor, phosphodiesterase 5A). Conversely, relative to placebo, colchicine significantly increased concentrations of eight molecules including secreted factors associated with metabolism and anti-thrombosis.ConclusionsIn adults with obesity, colchicine significantly affected concentrations of proteins involved in the innate immune system, endothelial function and atherosclerosis, uncovering new mechanisms behind its cardiometabolic effects. Further research is warranted to investigate whether colchicine’s IL-6 suppressive effects may be beneficial in COVID-19.

How to sample training/validation data is an important question for machine learning models, especially when the dataset is heterogeneous and skewed. In this paper, we propose a data sampling method that robustly selects training/validation data. We formulate the training/validation data sampling process as a two-player game: a trainer aims to sample training data so as to minimize the test error, while a validator adversarially samples validation data that can increase the test error. Robust sampling is achieved at the game equilibrium. To accelerate the searching process, we adopt reinforcement learning aided Monte Carlo trees search (MCTS). We apply our method to a car-following modeling problem, a complicated scenario with heterogeneous and random human driving behavior. Real-world data, the Next Generation SIMulation (NGSIM), is used to validate this method, and experiment results demonstrate the sampling robustness and thereby the model out-of-sample performance.

Task-oriented medical dialogue systems face two fundamental challenges: the explosion of state-action space caused by numerous diseases and symptoms and the sparsity of informative signals during interactive diagnosis. These issues significantly hinder the accuracy and efficiency of automated clinical reasoning. To address these problems, we propose MA-HRL, a multi-agent hierarchical reinforcement learning framework that decomposes the diagnostic task into specialized agents. A high-level controller coordinates symptom inquiry via multiple worker agents, each targeting a specific disease group, while a two-tier disease classifier refines diagnostic decisions through hierarchical probability reasoning. To combat sparse rewards, we design an information entropy-based reward function that encourages agents to acquire maximally informative symptoms. Additionally, medical knowledge graphs are integrated to guide decision-making and improve dialogue coherence. Experiments on the SymCat-derived SD dataset demonstrate that MA-HRL achieves substantial improvements over state-of-the-art baselines, including +7.2% diagnosis accuracy, +0.91% symptom hit rate, and +15.94% symptom recognition rate. Ablation studies further verify the effectiveness of each module. This work highlights the potential of hierarchical, knowledge-aware multi-agent systems for interpretable and scalable medical diagnosis.

Background Genetically engineered T cells have become an important therapy for B-cell malignancies. Measuring the efficiency of vector integration into the T cell genome is important for assessing the potency and safety of these cancer immunotherapies. Methods A digital droplet polymerase chain reaction (ddPCR) assay was developed and evaluated for assessing the average number of lenti- and retroviral vectors integrated into Chimeric Antigen Receptor (CAR) and T Cell Receptor (TCR)-engineered T cells. Results The ddPCR assay consistently measured the concentration of an empty vector in solution and the average number of CAR and TCR vectors integrated into T cell populations. There was a linear relationship between the average vector copy number per cell measured by ddPCR and the proportion of cells transduced as measured by flow cytometry. Similar vector copy number measurements were obtained by different staff using the ddPCR assay, highlighting the assays reproducibility among technicians. Analysis of fresh and cryopreserved CAR T and TCR engineered T cells yielded similar results. Conclusions ddPCR is a robust tool for accurate quantitation of average vector copy number in CAR and TCR engineered T cells. The assay is also applicable to other types of genetically engineered cells including Natural Killer cells and hematopoietic stem cells.

Background Chimeric antigen receptor (CAR) T-cells have demonstrated significant efficacy in targeting hematological malignancies, and their use continues to expand. Despite substantial efforts spent on the optimization of protocols for CAR T-cell manufacturing, critical parameters of cell culture such as pH or oxygenation are rarely actively monitored during cGMP CAR T-cell generation. A comprehensive understanding of the role that these factors play in manufacturing may help in optimizing patient-specific CAR T-cell therapy with maximum benefits and minimal toxicity. Methods This retrospective study examined cell culture supernatants from the manufacture of CAR T-cells for 20 patients with B-cell malignancies enrolled in a phase 1/2 clinical trial of anti-CD22 CAR T-cells. MetaFLEX was used to measure supernatant pH, oxygenation, and metabolites, and a Bio-Plex assay was used to assess protein levels. Correlations were assessed between the pH of cell culture media throughout manufacturing and cell proliferation as well as clinical outcomes. Next-generation sequencing was conducted to examine gene expression profiles of the final CAR T-cell products. Results A pH level at the lower range of normal at the beginning of the manufacturing process significantly correlated with measures of T-cell expansion and metabolism. Stable or rising pH during the manufacturing process was associated with clinical response, whereas a drop in pH was associated with non-response. Conclusions pH has potential to serve as an informative factor in predicting CAR T-cell quality and clinical outcomes. Thus, its active monitoring during manufacturing may ensure a more effective CAR T-cell product.