Single-cell multi-omic and spatial profiling of human kidneys implicates the fibrotic microenvironment in kidney disease progression.

Kidneys are intricate three-dimensional structures in the body, yet the spatial and molecular principles of kidney health and disease remain inadequately understood. We generated high-quality datasets for 81 samples, including single-cell, single-nuclear, spot-level (Visium) and single-cell resolution (CosMx) spatial-RNA expression and single-nuclear open chromatin, capturing cells from healthy, diabetic and hypertensive diseased human kidneys. Combining these data, we identify cell types and map them to their locations within the tissue. Unbiased deconvolution of the spatial data identifies the following four distinct microenvironments: glomerular, immune, tubule and fibrotic. We describe the complex organization of microenvironments in health and disease and find that the fibrotic microenvironment is able to molecularly classify human kidneys and offers an improved prognosis compared to traditional histopathology. We provide a comprehensive spatially resolved molecular roadmap of the human kidney and the fibrotic process, demonstrating the clinical utility of spatial transcriptomics.

DeepComBat: A statistically motivated, hyperparameter-robust, deep learning approach to harmonization of neuroimaging data.

Neuroimaging data acquired using multiple scanners or protocols are increasingly available. However, such data exhibit technical artifacts across batches which introduce confounding and decrease reproducibility. This is especially true when multi-batch data are analyzed using complex downstream models which are more likely to pick up on and implicitly incorporate batch-related information. Previously proposed image harmonization methods have sought to remove these batch effects; however, batch effects remain detectable in the data after applying these methods. We present DeepComBat, a deep learning harmonization method based on a conditional variational autoencoder and the ComBat method. DeepComBat combines the strengths of statistical and deep learning methods in order to account for the multivariate relationships between features while simultaneously relaxing strong assumptions made by previous deep learning harmonization methods. As a result, DeepComBat can perform multivariate harmonization while preserving data structure and avoiding the introduction of synthetic artifacts. We apply this method to cortical thickness measurements from a cognitive-aging cohort and show DeepComBat qualitatively and quantitatively outperforms existing methods in removing batch effects while preserving biological heterogeneity. Additionally, DeepComBat provides a new perspective for statistically motivated deep learning harmonization methods.

Spatial Single-cell Analysis Decodes Cortical Layer and Area Specification.

The human cerebral cortex, pivotal for advanced cognitive functions, is composed of six distinct layers and dozens of functionally specialized areas1,2. The layers and areas are distinguished both molecularly, by diverse neuronal and glial cell subtypes, and structurally, through intricate spatial organization3,4. While single-cell transcriptomics studies have advanced molecular characterization of human cortical development, a critical gap exists due to the loss of spatial context during cell dissociation5,6,7,8. Here, we utilized multiplexed error-robust fluorescence in situ hybridization (MERFISH)9, augmented with deep-learning-based cell segmentation, to examine the molecular, cellular, and cytoarchitectural development of human fetal cortex with spatially resolved single-cell resolution. Our extensive spatial atlas, encompassing 16 million single cells, spans eight cortical areas across four time points in the second and third trimesters. We uncovered an early establishment of the six-layer structure, identifiable in the laminar distribution of excitatory neuronal subtypes by mid-gestation, long before the emergence of cytoarchitectural layers. Notably, while anterior-posterior gradients of neuronal subtypes were generally observed in most cortical areas, a striking exception was the sharp molecular border between primary (V1) and secondary visual cortices (V2) at gestational week 20. Here we discovered an abrupt binary shift in neuronal subtype specification at the earliest stages, challenging the notion that continuous morphogen gradients dictate mid-gestation cortical arealization6,10. Moreover, integrating single-nuclei RNA-sequencing and in situ whole transcriptomics revealed an early upregulation of synaptogenesis in V1-specific Layer 4 neurons, suggesting a role of synaptogenesis in this discrete border formation. Collectively, our findings underscore the crucial role of spatial relationships in determining the molecular specification of cortical layers and areas. This work not only provides a valuable resource for the field, but also establishes a spatially resolved single-cell analysis paradigm that paves the way for a comprehensive developmental atlas of the human brain.

Optimizing EEG monitoring in critically ill children at risk for electroencephalographic seizures.

OBJECTIVE: Strategies are needed to optimally deploy continuous EEG monitoring (CEEG) for electroencephalographic seizure (ES) identification and management due to resource limitations. We aimed to construct an efficient multi-stage prediction model guiding CEEG utilization to identify ES in critically ill children using clinical and EEG covariates. METHODS: The largest prospective single-center cohort of 1399 consecutive children undergoing CEEG was analyzed. A four-stage model was developed and trained to predict whether a subject required additional CEEG at the conclusion of each stage given their risk of ES. Logistic regression, elastic net, random forest, and CatBoost served as candidate methods for each stage and were evaluated using cross validation. An optimal multi-stage model consisting of the top-performing stage-specific models was constructed. RESULTS: When evaluated on a test set, the optimal multi-stage model achieved a cumulative specificity of 0.197 and cumulative F1 score of 0.326 while maintaining a high minimum cumulative sensitivity of 0.938. Overall, 11 % of test subjects with ES were removed from the model due to a predicted low risk of ES (falsely negative subjects). CEEG utilization would be reduced by 32 % and 47 % compared to performing 24 and 48 h of CEEG in all test subjects, respectively. We developed a web application called EEGLE (EEG Length Estimator) that enables straightforward implementation of the model. CONCLUSIONS: Application of the optimal multi-stage ES prediction model could either reduce CEEG utilization for patients at lower risk of ES or promote CEEG resource reallocation to patients at higher risk for ES.