Shantou University

Data augmentation is crucial for addressing insufficient training data, especially for augmenting positive samples. However, existing methods mostly rely on neural network-based feedback for data augmentation and often overlook the optimization of feature distribution. In this study, we present a practical, distribution-preserving data augmentation pipeline that augments positive samples by optimizing a feature indicator (e.g., two-dimensional entropy), aiming to maintain alignment with the original data distribution. Inspired by the manifold hypothesis, we propose a Manifold Heuristic Optimization Algorithm (MHOA), which augments positive samples by exploring the low-dimensional Euclidean space around object contour pixels instead of the entire decision space. Guided by a "distribution-preservation-first" perspective, our approach explicitly optimizes fidelity to the original data manifold and only retains augmented samples whose feature statistics (e.g., mean, variance) align with the source class. It significantly improves image classification accuracy across neural networks, outperforming state-of-the-art data augmentation methods-especially when the dataset's feature indicator follows a Gaussian distribution. The algorithm's search space, focused on neighborhoods of key feature pixels, is the core driver of its superior performance.

Dinoflagellates and diatoms are key marine phytoplankton, with ecological roles strongly influenced by their associated phycosphere bacteria. However, the ecophysiological functions of these bacteria remain enigmatic as a result of insufficient taxonomic and genomic characterization. Here, by combining single-cell isolation with a custom statistical pipeline, we profiled resident bacterial communities associated with 108 diatom and 86 dinoflagellate strains, collected across temperate and tropical oceans worldwide. We examined genomic traits of key bacterial populations through whole-genome sequencing of representative isolates. Taxonomic compositions of dinoflagellate- and diatom-associated microbiota were distinct, highlighting host-specific differences. Each microbiota harboured characteristic genera with adaptive traits reflecting host metabolic profiles. Dinoflagellate-associated bacteria were enriched in genes responsible for motility and sulfur-compound use, whereas diatom-associated bacteria specialized in glycan use. We identified 'foundation' genera, defined as taxa with high occupancy and community-level impact in both phycosphere types (for example, Marivita and Marinobacter), sharing host-specific traits with characteristic bacteria while universally excelling in environmental response and resistance. Notably, foundation bacteria were enriched in Type VI secretion systems, emerging as a universal hallmark of phycosphere bacteria across global oceans. Overall, this study provides insights into the taxonomic and metabolic nature of phycosphere bacteria, highlighting the profound influences of interspecific interactions on marine ecological processes.

Growth differentiation factor 15 (GDF15) has emerged as a promising metabolic regulator with hepatoprotective properties in metabolic dysfunction-associated steatotic liver disease (MASLD), yet its underlying mechanisms remain elusive. Given that mitochondria are the primary site of fatty acid oxidation (FAO) and that mitochondrial morphodynamics are critical for normal hepatic lipid metabolism, we investigated how GDF15 regulates hepatic lipid homeostasis through mitochondrial dynamics. We established cellular steatosis models using primary rat hepatocytes exposed to lipotoxic palmitate (PA) or non-lipotoxic free fatty acid mixture (FFA, oleate: palmitate = 2: 1). Following GDF15 administration, we quantified lipid droplet content, expression of lipid metabolism genes, mitochondrial fatty acid translocation, and mitochondrial morphodynamics and function. The mechanistic role of ERK1/2 signalling was assessed through pharmacological inhibition. These findings were subsequently validated in adult progenitor cell-derived human liver organoids. GDF15 significantly mitigated both PA- and FFA-induced lipid accumulation by upregulating key FAO genes and down regulating lipid synthesis genes. Importantly, GDF15 corrected PA-induced mitochondrial fusion-fission imbalance by increasing mitochondrial fusion proteins MFN1 and OPA1 while modulating the activation of fission regulator DRP1. GDF15 enhanced fatty acid translocation into mitochondria and improved FAO. Mechanistically, GDF15 exerted these effects partially through inhibition of the ERK1/2 signalling pathway. Human liver organoid models further corroborated this protective mechanism of GDF15 against hepatic steatosis. Our study reveals that, specifically under lipotoxic conditions, GDF15 alleviates hepatocyte steatosis by preserving mitochondrial morphodynamics homeostasis and enhancing mitochondrial FAO capacity via ERK1/2 inhibition. These condition-specific mechanisms provide critical insights into GDF15's hepatoprotective effects and support its further investigation as a potential therapeutic target for MASLD.