Yunnan Investment Group (China)

Cell-cell communication regulates complex biological processes in multicellular systems. Existing scRNA-seq-based methods typically aggregate gene expression by clusters, overlooking within-cluster heterogeneity. We present scComm, a computational framework that infers cell-cell communications between individual cells using supervised contrastive learning. In simulations, scComm outperforms other methods and achieves up to 95% accuracy. Applied to colorectal cancer, it reveals cell-cell communications linked to PD-1 blockade response and tertiary lymphoid structures. In liver cancer, it identifies three novel tumor subtypes and angiogenesis-promoting neutrophil subtypes that have unique tumor microenvironments. scComm enables high-resolution cell-cell communication analysis, uncovering biological insights missed by existing approaches.

In this study, representative soil profiles developed on clastic rock parent materials in Yunnan Province were investigated to elucidate the formation mechanisms of soil magnetic properties under weakly magnetic parent material conditions and to evaluate the response of magnetic enhancement to chemical weathering and pedogenic differentiation. A combination of environmental magnetic measurements, bulk geochemical analyses, weathering index calculations, and ternary diagram discrimination was applied to characterize soil magnetic behavior, magnetic grain size distribution, and chemical weathering processes. The results show that the clastic rock parent materials exhibit overall low magnetic intensities, with low-frequency magnetic susceptibility (χlf) ranging from 2.543 × 10−8 m3/kg to 595.652 × 10−8 m3/kg. Under this weakly magnetic background, soils in the study area display pronounced pedogenic magnetic enhancement, with magnetic parameters showing clear and systematic vertical differentiation along soil profiles, indicating that soil magnetic signals are primarily controlled by pedogenesis. The frequency-dependent susceptibility (χfd%) generally falls within the range of 5.403%–17.574%, with a mean value of 12.898%, suggesting a substantial contribution from fine-grained magnetic particles. Magnetic grain size diagnostics further indicate that newly formed superparamagnetic (SP) and stable single-domain (SSD) particles generated during pedogenesis dominate the magnetic enhancement signal. The results of the Chemical Index of Alteration (CIA) indicate that approximately 78% of the profiles reach the strong weathering category (CIA > 85), while only 22% fall into the moderate weathering category (CIA: 65–85). Correlation analyses further reveal that grain-size-sensitive magnetic ratios (e.g., χfd%, χARM/SIRM) exhibit a strong correspondence with chemical weathering intensity indicators. These findings suggest that, under weakly magnetic parent material conditions, pedogenically induced magnetic enhancement can be more readily identified and quantitatively assessed. The integration of environmental magnetism and geochemical approaches, therefore, provides a robust framework for investigating pedogenic differentiation and supports high-resolution paleoenvironmental reconstruction in regions dominated by weakly magnetic parent materials.

Rapid growth of projects on high-fill sites under red clay and deep soft clayey foundations in Southwest China has exposed the limitations of conventional dynamic compaction in effective improvement depth and energy utilization. Punching and Squeezing Dynamic Compaction (PSDC) forms red clay–gravel composite piers through successive punching, backfilling and squeezing, offering potential advantages in deep densification; However, the mechanism of energy transmission and structural evolution remain unclear, constraining optimization of construction parameters and design. To address this gap, an integrated “Discrete element simulation–laboratory model testing–μCT 3D reconstruction” framework is established. Based on PFC3D with a Hertz contact model, impact-induced dynamic response and energy distribution were elucidated, and macro–meso consistency was verified against model tests and μCT-3D reconstructed piers, enabling systematic analysis of energy transfer, dissipation and skeleton reorganization under PSDC. Results show pronounced three-dimensional directional attenuation of impact energy: vertical transmission is the most efficient, the 45° oblique direction exhibits intermediate decay, and the horizontal direction attenuates rapidly with distance. Gravel content decisively governs energy pathways and skeletal architecture: a 60% gravel content produces continuous force chains, increases wave impedance, and concentrates energy at depth, promoting more effective compressive deformation and deep densification; in contrast, 50% gravel yields a more discrete skeleton, enhancing shallow random sliding, increasing sliding work, and promoting near-field dissipation. A directional attenuation model derived from a three-dimensional wavefront effectively fits the exponential decay of peak particle velocity with distance in the three directions and, for two representative gravel contents (50% and 60%), indicates a consistent chain linking gravel-skeleton connectivity, energy partitioning, and densification efficiency. These insights, obtained for 50%–60% gravel contents in high-fill red clay, illustrate how skeleton continuity regulates directional attenuation and densification, and they provide a basis for further extensions to broader mixture ratios and field scales.

Sediment accumulation in the drainage systems of mountain tunnels is a typical issue threatening operational safety. To explore the sedimentation behavior under the coupling of multiple factors, this study systematically analyzes the coupled effects of sediment content, flow rate, slope, and cross-sectional shape on sedimentation through full-scale experiments and numerical simulations. The results indicate that: (1) the sediment accumulation is linearly positively correlated with sediment concentration (fitting slope of 0.87) and exponentially negatively correlated with flow rate and slope (R2 > 0.90); (2) for drainage trenches with different cross-sectional shapes under the same boundary conditions, the maximum flow velocity and anti-sedimentation capacity rank as narrow rectangular > semi-circular ≈ inverted trapezoidal > rectangular; (3) the study proposes engineering anti-sedimentation strategies, such as moderately increasing the slope and adopting a periodic concentrated discharge model to enhance sediment transport capacity using peak flow; (4) under the premise of meeting drainage and flood control standards, the inverted trapezoidal or semi-circular cross-sections are preferred. The bottom waterway width can be reduced to increase flow velocity, thereby achieving a synergistic optimization of drainage efficiency and operational reliability. This provides a quantitative basis for the structural selection and anti-sedimentation design of tunnel drainage systems.

Unstable car-following behavior under truck-induced visual occlusion on mountainous two-lane roads significantly increases rear-end crash risk. However, compared with studies focusing on overtaking or curve risk prediction, the car-following-truck (CFT) risk and its influencing factors have received limited attention. Therefore, this study used unmanned aerial vehicles (UAVs) to collect high-resolution trajectory data of CFT scenarios on both straight and curved segments under truck-induced occlusion. First, the CFT risk was quantified based on an anticipated collision time (ACT) indicator, a two-dimensional surrogate safety measure that accounts for vehicle acceleration variations. Then, extreme value theory (EVT) was applied to calibrate alignment-specific risk thresholds. Finally, an XGBoost-based risk identification model was developed using vehicle dynamics-related features, and feature importance analysis combined with partial dependence interpretability was conducted to obtain key influencing factors. The results show that the calibrated ACT thresholds are approximately 3.838 s for straight segments and 4.385 s for curved segments, providing a reliable basis for risk classification. In addition, the XGBoost-based risk identification achieved accuracies of 90.63% and 95.87% for straight and curved segments, respectively. Further analysis indicates that CFT distance was the contributing factor. Moreover, risk increases markedly within a 10–20 m range on straight segments, while it rises rapidly once spacing falls below about 10 m on curved segments. Speed and acceleration differences exhibited stronger amplifying effects under short-spacing conditions. These findings provide a micro-behavioral basis for safety management and intelligent driving applications on mountainous roads with high truck mixing rates, supporting safer and more sustainable traffic operations.

To address the challenges of small object size, dense foliage, and strong background interference in tea leaf imagery under natural plantation environments, this study proposes an enhanced object detection model named CIS-YOLOv10. The model integrates three key optimizations: a Shape-IoU loss function for improved bounding box regression, the Convolutional Block Attention Module (CBAM) to enhance spatial focus, and StarNet to reduce computational complexity while maintaining accuracy. Experimental results demonstrate that CIS-YOLOv10 achieves Precision of 92.92%, Recall of 88.78%, F1-score of 90.80%, mAP of 93.46%, and an inference speed of 108.69 FPS—outperforming the baseline YOLOv10 in all metrics. The model offers robust performance, lightweight deployment, and strong application potential in intelligent tea picking scenarios.