Batch effects, which arise due to technical variations across different experimental factors, pose a significant challenge for single-cell transcriptomic analysis. Although various batch correction methods have been developed to mitigate these effects, they often indiscriminately mix data from different batches, leading to the removal of biologically meaningful signals. This limitation hinders comparative analyses across multiple conditions, an essential aspect of scientific research. Recent approaches attempt to address this issue by mapping data to separate spaces, but they prevent direct comparisons between conditions. Here, we propose Kanade, a batch correction method based on a variational autoencoder. Kanade explicitly disentangles batch effects from biological signals by specializing latent variables for different types of information. Using both simulated and real datasets, we demonstrate that Kanade selectively correct batch effects while preserving essential biological features, enabling more accurate comparative analyses in single-cell transcriptomics.

Background/Objectives: Accurate landmark identification underpins reliable cephalometric analysis. This study evaluated a ResNet50-based, single-stage regression convolutional neural network for direct automatic localization of 15 landmarks on lateral cephalograms. Methods: This retrospective study included 669 lateral cephalograms (669 patients): 619 for training and 50 randomly selected for an internal holdout test set. One of five orthodontists annotated each cephalogram, and coordinates served as the reference standard. One orthodontist independently re-annotated all test images after more than 2 weeks to assess reproducibility. Model performance was evaluated using Euclidean localization errors and success detection rates (SDR). Results: Across 750 landmark predictions, mean localization error was 1.25 ± 1.39 mm (95% confidence interval, 1.15–1.35 mm) and median error was 0.72 mm. SDRs within 1.0, 2.0, and 4.0 mm were 64.5%, 81.5%, and 94.4%, respectively. A statistically significant overall difference was observed among the 15 landmarks (Friedman χ²(14) = 28.84, p = 0.011). Point A had the numerically largest mean error (1.66 mm) and the mandibular central incisor the smallest (1.04 mm). The mean difference between original and repeated annotations was 0.52 ± 0.92 mm. Conclusions: In this single-center internal holdout test set, mean localization error was below the 2.0 mm benchmark. However, 18.5% of predictions exceeded 2.0 mm, and external validity remains unconfirmed. Artificial-intelligence-generated landmarks should be verified by orthodontists, and external validation is required before broader clinical use.

Accurate shell shape quantification is critical for studying biodiversity and evolution, yet intraspecific variability in bivalves makes morphology-based identification difficult. Traditional methods, including landmark-based analyses and elliptic Fourier descriptors, suffer either from subjectivity in homologous point selection or from limited use of contour information. Here, we introduce Morpho-VAE, a deep generative framework integrating a variational autoencoder with a supervised classifier, to analyze shell images of five Anadara species. Morpho-VAE outperforms conventional approaches in species classification by embedding morphological variation into a low-dimensional space where species cluster distinctly. To highlight species-specific morphological patterns, we develop a patch masking assay, revealing the hinge line as a shared morphological marker across species and species-specific regions near the umbo and anterior ventral margin. The decoder further enables morphological visualization via image reconstruction and interpolation. Our results show that Morpho-VAE can automatically extract species-defining morphological patterns from raw images, providing complementary or novel insights beyond traditional morphometric methods.

People do not always feel the way they appear. Someone who seems emotionally stable may be struggling internally, whereas someone who appears distressed may experience their state very differently. This gap matters in psychiatry, where assessment often relies on symptom scales and external evaluation. Here we developed a hierarchical variational Bayesian framework to jointly estimate population-derived dynamics, reflecting an external reference based on pooled data, and individual-derived dynamics, reflecting each participant's subjective symptom organization. We applied this framework to time-series PHQ-9 data of depressive states from 248 participants during the COVID-19 pandemic. The population landscape contained three major stable states, whereas individualized landscapes often diverged substantially from this shared structure. We quantified this mismatch as individual-group landscape divergence, which was associated not only with depressive severity but also with modern-type depression-related traits (TACS-22) and interpersonal sensitivity-self traits (IPS-22). These findings reveal a previously unquantified gap between population-level and individual-level symptom organization.