Uncertainty-aware Health Diagnostics via Class-balanced Evidential Deep Learning.

Uncertainty quantification is critical for ensuring the safety of deep learning-enabled health diagnostics, as it helps the model account for unknown factors and reduces the risk of misdiagnosis. However, existing uncertainty quantification studies often overlook the significant issue of class imbalance, which is common in medical data. In this paper, we propose a class-balanced evidential deep learning framework to achieve fair and reliable uncertainty estimates for health diagnostic models. This framework advances the state-of-the-art uncertainty quantification method of evidential deep learning with two novel mechanisms to address the challenges posed by class imbalance. Specifically, we introduce a pooling loss that enables the model to learn less biased evidence among classes and a learnable prior to regularize the posterior distribution that accounts for the quality of uncertainty estimates. Extensive experiments using benchmark data with varying degrees of imbalance and various naturally imbalanced health data demonstrate the effectiveness and superiority of our method. Our work pushes the envelope of uncertainty quantification from theoretical studies to realistic healthcare application scenarios. By enhancing uncertainty estimation for class-imbalanced data, we contribute to the development of more reliable and practical deep learning-enabled health diagnostic systems1.

Expedient and divergent synthesis of unnatural peptides through cobalt-catalyzed diastereoselective umpolung hydrogenation.

The development of a reliable method for asymmetric synthesis of unnatural peptides is highly desirable and particularly challenging. In this study, we present a versatile and efficient approach that uses cobalt-catalyzed diastereoselective umpolung hydrogenation to access noncanonical aryl alanine peptides. This protocol demonstrates good tolerance toward various functional groups, amino acid sequences, and peptide lengths. Moreover, the versatility of this reaction is illustrated by its successful application in the late-stage functionalization and formal synthesis of various representative chiral natural products and pharmaceutical scaffolds. This strategy eliminates the need for synthesizing chiral noncanonical aryl alanines before peptide formation, and the hydrogenation reaction does not result in racemization or epimerization. The underlying mechanism was extensively explored through deuterium labeling, control experiments, HRMS identification, and UV-Vis spectroscopy, which supported a reasonable CoI/CoIII catalytic cycle. Notably, acetic acid and methanol serve as safe and cost-effective hydrogen sources, while indium powder acts as the terminal electron source.

Human-centred artificial intelligence for mobile health sensing: challenges and opportunities.

Advances in wearable sensing and mobile computing have enabled the collection of health and well-being data outside of traditional laboratory and hospital settings, paving the way for a new era of mobile health. Meanwhile, artificial intelligence (AI) has made significant strides in various domains, demonstrating its potential to revolutionize healthcare. Devices can now diagnose diseases, predict heart irregularities and unlock the full potential of human cognition. However, the application of machine learning (ML) to mobile health sensing poses unique challenges due to noisy sensor measurements, high-dimensional data, sparse and irregular time series, heterogeneity in data, privacy concerns and resource constraints. Despite the recognition of the value of mobile sensing, leveraging these datasets has lagged behind other areas of ML. Furthermore, obtaining quality annotations and ground truth for such data is often expensive or impractical. While recent large-scale longitudinal studies have shown promise in leveraging wearable sensor data for health monitoring and prediction, they also introduce new challenges for data modelling. This paper explores the challenges and opportunities of human-centred AI for mobile health, focusing on key sensing modalities such as audio, location and activity tracking. We discuss the limitations of current approaches and propose potential solutions.

Evaluating Listening Performance for COVID-19 Detection by Clinicians and Machine Learning: Comparative Study.

BACKGROUND: To date, performance comparisons between men and machines have been carried out in many health domains. Yet machine learning (ML) models and human performance comparisons in audio-based respiratory diagnosis remain largely unexplored. OBJECTIVE: The primary objective of this study was to compare human clinicians and an ML model in predicting COVID-19 from respiratory sound recordings. METHODS: In this study, we compared human clinicians and an ML model in predicting COVID-19 from respiratory sound recordings. Prediction performance on 24 audio samples (12 tested positive) made by 36 clinicians with experience in treating COVID-19 or other respiratory illnesses was compared with predictions made by an ML model trained on 1162 samples. Each sample consisted of voice, cough, and breathing sound recordings from 1 subject, and the length of each sample was around 20 seconds. We also investigated whether combining the predictions of the model and human experts could further enhance the performance in terms of both accuracy and confidence. RESULTS: The ML model outperformed the clinicians, yielding a sensitivity of 0.75 and a specificity of 0.83, whereas the best performance achieved by the clinicians was 0.67 in terms of sensitivity and 0.75 in terms of specificity. Integrating the clinicians' and the model's predictions, however, could enhance performance further, achieving a sensitivity of 0.83 and a specificity of 0.92. CONCLUSIONS: Our findings suggest that the clinicians and the ML model could make better clinical decisions via a cooperative approach and achieve higher confidence in audio-based respiratory diagnosis.

Exploring Longitudinal Cough, Breath, and Voice Data for COVID-19 Progression Prediction via Sequential Deep Learning: Model Development and Validation.

BACKGROUND: Recent work has shown the potential of using audio data (eg, cough, breathing, and voice) in the screening for COVID-19. However, these approaches only focus on one-off detection and detect the infection, given the current audio sample, but do not monitor disease progression in COVID-19. Limited exploration has been put forward to continuously monitor COVID-19 progression, especially recovery, through longitudinal audio data. Tracking disease progression characteristics and patterns of recovery could bring insights and lead to more timely treatment or treatment adjustment, as well as better resource management in health care systems. OBJECTIVE: The primary objective of this study is to explore the potential of longitudinal audio samples over time for COVID-19 progression prediction and, especially, recovery trend prediction using sequential deep learning techniques. METHODS: Crowdsourced respiratory audio data, including breathing, cough, and voice samples, from 212 individuals over 5-385 days were analyzed, alongside their self-reported COVID-19 test results. We developed and validated a deep learning-enabled tracking tool using gated recurrent units (GRUs) to detect COVID-19 progression by exploring the audio dynamics of the individuals' historical audio biomarkers. The investigation comprised 2 parts: (1) COVID-19 detection in terms of positive and negative (healthy) tests using sequential audio signals, which was primarily assessed in terms of the area under the receiver operating characteristic curve (AUROC), sensitivity, and specificity, with 95% CIs, and (2) longitudinal disease progression prediction over time in terms of probability of positive tests, which was evaluated using the correlation between the predicted probability trajectory and self-reported labels. RESULTS: We first explored the benefits of capturing longitudinal dynamics of audio biomarkers for COVID-19 detection. The strong performance, yielding an AUROC of 0.79, a sensitivity of 0.75, and a specificity of 0.71 supported the effectiveness of the approach compared to methods that do not leverage longitudinal dynamics. We further examined the predicted disease progression trajectory, which displayed high consistency with longitudinal test results with a correlation of 0.75 in the test cohort and 0.86 in a subset of the test cohort with 12 (57.1%) of 21 COVID-19-positive participants who reported disease recovery. Our findings suggest that monitoring COVID-19 evolution via longitudinal audio data has potential in the tracking of individuals' disease progression and recovery. CONCLUSIONS: An audio-based COVID-19 progression monitoring system was developed using deep learning techniques, with strong performance showing high consistency between the predicted trajectory and the test results over time, especially for recovery trend predictions. This has good potential in the postpeak and postpandemic era that can help guide medical treatment and optimize hospital resource allocations. The changes in longitudinal audio samples, referred to as audio dynamics, are associated with COVID-19 progression; thus, modeling the audio dynamics can potentially capture the underlying disease progression process and further aid COVID-19 progression prediction. This framework provides a flexible, affordable, and timely tool for COVID-19 tracking, and more importantly, it also provides a proof of concept of how telemonitoring could be applicable to respiratory diseases monitoring, in general.

Characterization and mutagenesis of a novel Mycobacterium smegmatis-derived glutamate decarboxylase active at neutral pH.

γ-aminobutyric acid (GABA) has various physiological functions and is widely used in medicine, food, and other fields. Glutamate decarboxylase (GAD) is a key enzyme that catalyzes the decarboxylation of L-glutamate to synthesize GABA. However, the industrial application of microorganism-derived GAD is limited by its rapid loss of enzymatic activity with pH approaching neutrality. In this study, a novel glutamate decarboxylase, GADMSM, from Mycobacterium smegmatis was overexpressed and purified. On the basis of homologous modeling and substrate molecular docking, several GADMSM mutants were constructed, and their enzymatic properties were analyzed. The results showed that the optimal pH of wild-type GADMSM is 5.4; at pH 6.2, 22.8% enzymatic activity was retained. The T211I replacement in GAD and C-terminal deletion mutant GADMSMΔC showed relatively high catalytic activity in a pH range of 5.0-7.0. The Vmax and Km values of GADMSMΔC were 14.69 and 5.70, respectively, at pH 5.5, and 9.87 and 6.17, respectively, at pH 7.0. Compared with the wild-type GAD, GADMSMΔC maintained higher affinity and enzymatic activity of the substrate, maintaining 78.5% of the highest enzymatic activity even at pH 7.0, which is the highest reported activity retention for GAD under neutral pH condition. Therefore, GADMSMΔC can be used for the transformation of high-yielding strains and industrial production of GABA.

Sounds of COVID-19: exploring realistic performance of audio-based digital testing.

To identify Coronavirus disease (COVID-19) cases efficiently, affordably, and at scale, recent work has shown how audio (including cough, breathing and voice) based approaches can be used for testing. However, there is a lack of exploration of how biases and methodological decisions impact these tools' performance in practice. In this paper, we explore the realistic performance of audio-based digital testing of COVID-19. To investigate this, we collected a large crowdsourced respiratory audio dataset through a mobile app, alongside symptoms and COVID-19 test results. Within the collected dataset, we selected 5240 samples from 2478 English-speaking participants and split them into participant-independent sets for model development and validation. In addition to controlling the language, we also balanced demographics for model training to avoid potential acoustic bias. We used these audio samples to construct an audio-based COVID-19 prediction model. The unbiased model took features extracted from breathing, coughs and voice signals as predictors and yielded an AUC-ROC of 0.71 (95% CI: 0.65-0.77). We further explored several scenarios with different types of unbalanced data distributions to demonstrate how biases and participant splits affect the performance. With these different, but less appropriate, evaluation strategies, the performance could be overestimated, reaching an AUC up to 0.90 (95% CI: 0.85-0.95) in some circumstances. We found that an unrealistic experimental setting can result in misleading, sometimes over-optimistic, performance. Instead, we reported complete and reliable results on crowd-sourced data, which would allow medical professionals and policy makers to accurately assess the value of this technology and facilitate its deployment.

Multiregional profiling of the brain transmembrane proteome uncovers novel regulators of depression.

Transmembrane proteins play vital roles in mediating synaptic transmission, plasticity, and homeostasis in the brain. However, these proteins, especially the G protein-coupled receptors (GPCRs), are underrepresented in most large-scale proteomic surveys. Here, we present a new proteomic approach aided by deep learning models for comprehensive profiling of transmembrane protein families in multiple mouse brain regions. Our multiregional proteome profiling highlights the considerable discrepancy between messenger RNA and protein distribution, especially for region-enriched GPCRs, and predicts an endogenous GPCR interaction network in the brain. Furthermore, our new approach reveals the transmembrane proteome remodeling landscape in the brain of a mouse depression model, which led to the identification of two previously unknown GPCR regulators of depressive-like behaviors. Our study provides an enabling technology and rich data resource to expand the understanding of transmembrane proteome organization and dynamics in the brain and accelerate the discovery of potential therapeutic targets for depression treatment.

[Effects of fertilization and planting patterns on soil aggregate and carbon distribution in farmland of the Loess Plateau, Northwest China]. / ç§æ¤æ¨¡å¼å’Œæ–½è‚¥å¯¹é»„åœŸæ—±å¡¬å†œç”°åœŸå£¤å›¢èšä½“åŠå ¶ç¢³åˆ†å¸ƒçš„å½±å“.

Based on a long-term experiment in the Changwu Agro-ecological Experimental Station in Xianyang, Shaanxi, China, we examined the effects of fertilization and planting patterns on soil aggregate quantity, aggregate stability and total carbon and organic carbon distribution in different aggregate fractions through dry and wet sieving methods, as well as the TOC combustion method. There were ten treatments, including uncultivated (R), wheat continuous cropping (CK/W), wheat-corn rotation (L), and nitrogen fertilizer (N), phosphorus fertilizer (P), nitrogen and phosphorus fertilizer (NP), organic fertilizer (M), nitrogen and organic fertilizer (NM), phosphorus and organic fertilizer (PM), nitrogen and phosphorus and organic fertilizer (NPM) for CK/W. The results showed that fertilizer application and planting patterns affected soil aggregate distribution and stability, the contents and contribution rates of total C and organic C. Force-stable aggregate was mainly constituted by >0.25 mm aggregate (>67%), which was reduced by fertilization. Continuous cropping decreased micro-aggregate while rotation facilitated it and the effect was larger than fertilization. Water-stable aggregate was mainly comprised of micro-aggregate (<0.25 mm), the contribution of which was larger than 61%. Both fertilizer application and planting pattern reduced water-stable micro-aggregate. Fertilizer application and planting pattern decreased the percentage of aggregate destruction rate (PAD) and increased macro-aggregate (>0.25 mm, R0.25) content. Organic fertilizer significantly improved total C and organic C concentrations in all the fractions of force-stable aggregates. Continuous cropping and rotation cropping increased total C concentration in all the aggregate fractions while rotation cropping significantly decreased organic C concentration. Single N and P fertilization decreased soil total C concentration, while mixed application of N and P fertilizers, and organic fertilizer significantly increased soil total C concentration. The effect of planting patterns on soil total C was lower than that of fertilization. Both continuous cropping and rotation cropping increased soil total C. Mixed application of N and P fertilizers, and organic fertilizer signifi-cantly increased soil organic C concentration while single N and P fertilization decreased it. The effect of planting patterns on soil organic C was lower than that of fertilization, while rotation cropping did not facilitate soil organic C. Micro-aggregate was the most notable size fraction to total carbon and organic C, with the contribution being 21.2%-33.6%. Fertilization and planting pattern increased the contribution rate of micro-aggregate in soil total C. NP and NPM significantly increased the contribution rate of micro-aggregate in soil total C and soil organic C. The effect of rotation cropping was most obvious in driving the contribution rate of micro-aggregate in soil total C and soil organic C.

[Effects of long-term fertilization on soil organic phosphorus fractions and wheat yield in farmland of Loess Plateau]. / 长期施肥对黄土旱塬农田土壤有机磷组分及小麦产量的影响.

Understanding the effects of long-term fertilization on soil organic phosphorus fractions and wheat yield in the Loess Plateau can provide theoretical support for improving phosphorus conversion, utilization, and rational use of fertilizer. We examined the effects of different fertilizer treatments on soil organic phosphorus fractions, wheat yield and soil properties of a farmland in the long-term (1984-2016) positioning test station of Changwu loess soil. There were eight treatments, including no fertilizer (CK), single application of nitrogen fertilizer (N), single application of phosphorus fertilizer (P), application of nitrogen and phosphorus fertilizer (NP), single application of organic fertilizer (M), nitrogen combined with organic fertilizer (MN), phosphorus combined with organic fertilizer (MP), nitrogen and phosphorus combined with organic fertilizer (MNP). The results showed that the range of soil organic phosphorus content was 244.7-429.1 mg·kg-1 after long-term fertilization. Except for the N treatment, organic phosphorus content was significantly increased by 15.4%-47.9% compared to CK. Long-term application of phosphorus fertilizer changed the content of organic phosphorus fractions in the surface soil (0-20 cm). The treatments of MP and MNP significantly increased the contents of labile organic phosphorus and moderately labile organic phosphorus. The treatments of N, P and NP significantly reduced the content of moderately stable organic phosphorus. The treatments of N, P, NP, MN, MP, MNP all significantly increased the highly stable organic phosphorus. The ratio of soil organic phosphorus fractions to total organic phosphorus content was in order of moderately labile organic phosphorus > highly stable organic phosphorus > labile organic phosphorus > moderately stable organic phosphorus. After long-term fertilizer application, the combination of nitrogen and phosphorus fertilizers, especially with organic fertilizers, significantly increased wheat biomass yield and grain yield. Among all the examined soil properties, organic matter, Olsen-P and total inorganic phosphorus were significantly positively correlated with wheat yield. MP and M could significantly increase the content of Olsen-P, total phosphorus, total inorganic phosphorus, labile organic phosphorus and moderately labile organic phosphorus in the loess soil of Loess Plateau. Our results indicated that the organic and phosphorus fertilizers could improve soil phosphorus components that could be more easily absorbed by crops. In summary, the combination of nitrogen and phosphorus fertilizers, especially with organic fertilizers, could increase soil phosphorus supply in the region and promote the wheat yield, which is important for improving soil quality in the Loess Plateau.

The critical role of ASD-related gene CNTNAP3 in regulating synaptic development and social behavior in mice.

Accumulated genetic evidences indicate that the contactin associated protein-like (CNTNAP) family is implicated in autism spectrum disorders (ASD). In this study, we identified genetic mutations in the CNTNAP3 gene from Chinese Han ASD cohorts and Simons Simplex Collections. We found that CNTNAP3 interacted with synaptic adhesion proteins Neuroligin1 and Neuroligin2, as well as scaffolding proteins PSD95 and Gephyrin. Significantly, we found that CNTNAP3 played an opposite role in controlling the development of excitatory and inhibitory synapses in vitro and in vivo, in which ASD mutants exhibited loss-of-function effects. In this study, we showed that the male Cntnap3-null mice exhibited deficits in social interaction， spatial learning and prominent repetitive behaviors. These evidences elucidate the pivotal role of CNTNAP3 in synapse development and social behaviors, providing mechanistic insights into ASD.

Sesquiterpenoids with diverse carbon skeletons from the roots of Cichorium glandulosum and their anti-inflammatory activities.

A total of thirteen sesquiterpenoids with diverse skeletons including four new sesquiterpenoids, glandulosines A - D (1-4), a new natural product, glandulosine E (5), and eight known sesquiterpene lactones (6-13) were isolated from the roots of Cichorium glandulosum Boiss. et Huet (Asteraceae). Their structures were determined by extensive spectroscopic experiments including NMR, electronic circular dichroism (ECD), calculated ECD, Rh2(OCOCF3)4-induced ECD, and single-crystal X-ray diffraction analysis, as well as chemical methods. This is the first report of the crystal structure of 11ß,13-dihydrolactucin (11). Thirteen isolated sesquiterpenoids (1-13) were evaluated for their anti-inflammatory activities in vitro, and three guaiane sesquiterpene lactones, glandulosine E (5), scorzoside (9), and lactucin (10) showed moderate inhibitory activity against LPS-induced nitric oxide (NO) production in RAW 264.7 macrophages.

[Effects of long-term fertilization on soil microbial abundance in farmland of the Loess Plateau, China]. / 长期施肥对黄土旱塬农田土壤微生物丰度的影响.

Based on a long-term experiment in the Changwu Agro-ecological Experimental Station, the real-time PCR analysis was used to examine the soil microbial abundance and to reveal the effects on soil microbial community under different long-term fertilization systems. The results showed that compared to the CK (barren field), the population of bacteria increased by 21% and archaea by 32% in treatment with inorganic fertilizer application. The abundance of bacteria in the treatment of chemical fertilizer combined with organic fertilizer increased by 37% and archaea by 36%. The treatment with chemical fertilizer mixed with organic fertilizer significantly increased the abundance of bacteria and archaea. The soil AOB increased by 7.13 times while the soil AOA only by 0.2 folds after 30-year application of chemical nitrogen fertilizer. AOB was highly responsive to fertilizer application, especially to nitrogen fertilizer. Compared with the single nitrogen application and the application of nitrogen fertilizer mixed with organic fertilizer, phosphorus fertilizer significantly increased the abundance of nifH and pmoA. The content of nifH, nirS cd and pmoA in the abandoned land was significantly higher than that in the cultivated soil. Results from the correlation analysis on soil basic physical and chemical properties indicated that soil pH, total nitrogen and organic carbon were key factors affecting soil microbial community abundance. In conclusion, long-term fertilization significantly changed soil microbial abundance, and fertilization patterns and cultivating methods had significant effect on microbial community abundance.

Mollebenzylanols A and B, Highly Modified and Functionalized Diterpenoids with a 9-Benzyl-8,10-dioxatricyclo[5.2.1.01,5]decane Core from Rhododendron molle.

Two highly modified and functionalized diterpenoids, mollebenzylanols A (1) and B (2), and a known grayanane diterpenoid rhodojaponin III (3) were isolated from Rhododendron molle. Their structures were determined by spectroscopic data analysis, an electronic circular dichroism (ECD) exciton chirality method, ECD calculations, and X-ray diffraction analysis of the p-bromobenzoate ester of 1 (1a). Compounds 1 and 2 possess an unprecedented diterpene carbon skeleton featuring a unique 9-benzyl-8,10-dioxatricyclo[5.2.1.01,5]decane core, and their plausible biosynthetic pathways are proposed. Their PTP1B inhibitory activity and modes of action were investigated.

[Distribution characteristics of soil organic carbon under different forest restoration modes on opencast coal mine dump].

The content and storage of soil organic carbon (SOC) were compared in six wood restoration modes and adjacent abandoned land on opencast coal mine dump, and the mechanisms behind the differences and their influencing factors were analyzed. Results showed that the contents of SOC in six wood lands were significantly higher (23.8%-53.2%) than that of abandoned land (1.92 g · kg⁻¹) at 0-10 cm soil depth, the index were significantly higher (5.8%-70.4%) at 10-20 cm soil depth than the abandoned land (1.39 g · kg⁻¹), and then the difference of the contents of SOC in the deep soil (20-100 cm) were not significant. The contents of SOC decreased with increase of soil depth, but the decreasing magnitude of the topsoil (0-20 cm) was higher than that of the deep soil (20-100 cm). Compared with the deep soil, the topsoil significant higer storage of SOC in different woods, the SOC storage decreased with the soil depth. Along the 0-100 cm soil layer, the storage of SOC in six wood lands higher (18.1%-42.4%) than that of the abandoned land (17.52 t · hm⁻²). The SOC storage of Amorpha fruticosa land (24.95 t · hm⁻²) was obviously higher than that in the other wood lands. The SOC storage in the shrub lands was 12.4% higher than that of the arbor woods. There were significantly positive correlations among forest litter, fine root biomass, soil water content and SOC on the dump. Consequently, different plantation restorations significantly improved the SOC level on the dump in 0-100 cm soil, especially the topsoil. But there was still a big gap about SOC level between the wood restoration lands and the original landform. To improve the SOC on opencast coal mine dump, A. fruticosa could be selected as the main wood vegetation.

[Soil organic carbon storage changes with land reclamation under vegetation reconstruction on opencast coal mine dump].

Vegetation reconstruction was an effective solution to reclaim the opencast coal mine dump which was formed in the process of mining. To understand the impact of the vegetation reconstruction patterns' on the mine soil organic carbon (SOC) storage was essential for selecting the methods of vegetation restoration and also important for accurately estimating the potential of the soil carbon sequestration. The study area was on the Heidaigou opencast coal mine, which was 15 years reclaimed coal mine dump in Zhungeer, Inner Mongolia autonomous region, we selected 5 vegetation reconstruction patterns (natural recovery land, grassland, bush land, mixed woodland of arbor and bush, arbor land), and 16 vegetation types, 408 soil samples (0-100 m), to study the effect of the vegetation reconstruction patterns on the SOC storage. The results were showed as follows: (1) on the reclaimed coal mine dump, the vegetation reconstruction patterns significantly affected the SOC content and its distribution in the soil profile (P < 0.05). The surface 0-10 cm SOC content was grassland > shrub land > arbor forest > mixed forest of arbor and shrub > natural recovery land, in which the grassland, shrub land and arbor forest were about 2.2, 1.3, and 1.3 times of natural recovery land (2.14 g · kg(-1)) respectively. The total nitrogen (TN) showed the similar trends. (2) Among the vegetation types, Medicago sativa had the highest surface SOC content (5.71 g · kg(-1)) and TN content (0.49 g · kg(-1)), that were 171.3% and 166.7% higher than the natural recovery land, and two times of Hippophae rhamnoides, Amorpha fruticosa + Pinus tabulaeformis and Robinia pseudoacacia. (3) The effect of vegetation types on SOC mainly concentrated in the 0-20 cm depth, and the effect on TN accounted for 40 cm. (4) For the SOC storage, the order was original landform area > reclaimed dump > new dump and grassland > woodland (including arbor and shrub land). After 15 years revegetation, the soil carbon storage of the grassland, shrub land and arbor land were increased by 15.47 t · hm(-2), 6.93 t · hm(-2) and 6.95 t · hm(-2) respectively in the 100 cm depth, which were equivalent to 2/3, 1/2 and 1/2 of the original landform levels. The results showed a great ability of carbon sequestration.