Search for: All records

Histopathology images capture tissue morphology, while spatial transcriptomics (ST) provides spatially resolved gene expression, offering complementary molecular insights. However, acquiring ST data is costly and time-consuming, limiting its practical use. To address this, we propose HAGE (Hierarchical Alignment Gene-Enhanced), a framework that enhances pathology representation learning by predicting gene expression directly from histological images and integrating molecular context into the pathology model. HAGE leverages gene-type embeddings, which encode relationships among genes, guiding the model in learning biologically meaningful expression patterns. To further improve alignment between histology and gene expression, we introduce a hierarchical clustering strategy that groups image patches based on molecular and visual similarity, capturing both local and global dependencies. HAGE consistently outperforms existing methods across six datasets. In particular, on the HER2+ breast cancer cohort, it significantly improves the Pearson correlation coefficient by 8.0% and achieves substantial reductions in mean squared error and mean absolute error by 18.1% and 38.0%, respectively. Beyond gene expression prediction, HAGE improves downstream tasks, such as patch-level cancer classification and whole-slide image diagnostics, demonstrating its broader applicability. To the best of our knowledge, HAGE is the first framework to integrate gene co-expression as prior knowledge into a pathology image encoder via a cross-attention mechanism, enabling more biologically informed and accurate pathology representations. https://github.com/uta-smile/gene_expression. 

Exploring the functions of genes and gene products is crucial to a wide range of fields, including medical research, evolutionary biology, and environmental science. However, discovering new functions largely relies on expensive and exhaustive wet lab experiments. Existing methods of automatic function annotation or prediction mainly focus on protein function prediction with sequence, 3D-structures or protein family information. In this study, we propose to tackle the gene function prediction problem by exploring Gene Ontology graph and annotation with BERT (GoBERT) to decipher the underlying relationships among gene functions. Our proposed novel function prediction task utilizes existing functions as inputs and generalizes the function prediction to gene and gene products. Specifically, two pre-train tasks are designed to jointly train GoBERT to capture both explicit and implicit relations of functions. Neighborhood prediction is a self-supervised multi-label classification task that captures the explicit function relations. Specified masking and recovering task helps GoBERT in finding implicit patterns among functions. The pre-trained GoBERT possess the ability to predict novel functions for various gene and gene products based on known functional annotations. Extensive experiments, biological case studies, and ablation studies are conducted to demonstrate the superiority of our proposed GoBERT. 

Abstract El Niño–Southern Oscillation (ENSO), the dominant mode of interannual variability in the tropical Pacific, is well known to affect the extratropical climate via atmospheric teleconnections. Extratropical atmospheric variability may in turn influence the occurrence of ENSO events. The winter North Pacific Oscillation (NPO), as the secondary dominant mode of atmospheric variability over the North Pacific, has been recognized as a potential precursor for ENSO development. This study demonstrates that the preexisting winter NPO signal is primarily excited by sea surface temperature (SST) anomalies in the equatorial western–central Pacific. During ENSO years with a preceding winter NPO signal, which accounts for approximately 60% of ENSO events observed in 1979–2021, significant SST anomalies emerge in the equatorial western–central Pacific in the preceding autumn and winter. The concurrent presence of local convection anomalies can act as a catalyst for NPO-like atmospheric circulation anomalies. In contrast, during other ENSO years, significant SST anomalies are not observed in the equatorial western–central Pacific during the preceding winter, and correspondingly, the NPO signal is absent. Ensemble simulations using an atmospheric general circulation model driven by observed SST anomalies in the tropical western–central Pacific can well reproduce the interannual variability of observed NPO. Therefore, an alternative explanation for the observed NPO–ENSO relationship is that the preceding winter NPO is a companion to ENSO development, driven by the precursory SST signal in the equatorial western–central Pacific. Our results suggest that the lagged relationship between ENSO and the NPO involves a tropical–extratropical two-way coupling rather than a purely stochastic forcing of the extratropical atmosphere on ENSO. 

While the prominent influence of El Niño‐Southern Oscillation (ENSO) on the Indian Ocean Oscillation (IOD) is widely recognized, intricate relationships between them are often invoked that introduce challenges into seasonal predictions. Previous studies have shown that different flavors of El Niño exhibit distinct associations with the IOD. In this study, we demonstrate that La Niña's teleconnection to the IOD is primarily controlled by its longitudinal position. Westward‐displaced La Niña events tend to produce stronger negative convection anomalies in the central Pacific and more pronounced Walk Circulation anomalies, thereby triggering strong negative IOD events. In contrast, eastward‐displaced La Niña events are usually accompanied by feeble convection response due to the excessively cold conditions in the cold tongue, yielding insignificant IOD response. The pivotal role of La Niña's longitudinal position on the IOD's response is realistically reproduced by targeted pacemaker experiments, providing new insights into inter‐basin climate connections. 

Abstract Understanding how the tropical Pacific responds to rising greenhouse gases in recent decades is of paramount importance given its central role in global climate systems. Extensive research has explored the long-term trends of tropical Pacific sea surface temperatures (SSTs) and the overlying atmosphere, yet the historical change in the upper ocean has received far less attention. Here, we present compelling evidence of a prominent subsurface cooling pattern along the thermocline in the central-to-eastern tropical Pacific since 1958. This subsurface cooling has been argued to be contributing to the observed cooling or lack of warming of the equatorial cold tongue SST. We further demonstrate that different mechanisms are responsible for different parts of the subsurface cooling. In the central-to-eastern equatorial Pacific and the southeastern off-equatorial Pacific, where zonal wind stress strengthens, a pronounced subsurface cooling trend emerges just above the thermocline that is closely tied to increased Ekman pumping. In the eastern equatorial Pacific where zonal wind stress weakens, the westward surface current and eastward Equatorial Undercurrent weaken as well, resulting in reduced vertical current shear and increased ocean stability, which suppresses vertical mixing and leads to local cooling. We conclude that the historical subsurface cooling is primarily linked to dynamical adjustments of ocean currents to tropical surface wind stress changes. 

Abstract The Pacific Meridional Mode (PMM) has long been associated with extra‐tropical air‐sea coupling processes, which are thought to influence the development of El Niño‐Southern Oscillation (ENSO). Here we show that the PMM on seasonal to interannual timescales is closely associated with a newly proposed tropical mode known as the ENSO Combination mode (C‐mode), which arises from the nonlinear interaction between ENSO and the background annual cycle in the deep tropics. The PMM exhibits a remarkable resemblance with the C‐mode in atmospheric patterns, spectral characteristics, and local impacts. Based on a simple Hasselmann‐type model, we further demonstrate that the C‐mode‐related atmospheric anomalies can effectively drive PMM‐like sea surface temperature anomalies. As the C‐mode captures seasonally modulated ENSO characteristics, the seasonal‐to‐interannual PMM variability could naturally establish a connection with ENSO, thereby offering an alternative explanation for the observed relationship between PMM and ENSO. 

Abstract The equatorial cold tongue region has not warmed up in response to historical radiative forcing in the real world, contrary to the strong warming often simulated by climate models. Here we demonstrate that climate models fail to represent one or both of the key processes driving observed sea surface temperature (SST) pattern formation: a realistic surface wind stress pattern shaping subsurface cooling through wind‐driven circulation changes, and effective connectivity between subsurface and surface temperatures via upwelling and mixing. Consequently, none of the models approximate the observed lack of cold tongue SST warming and strengthening of zonal SST gradient across the equatorial Pacific. Furthermore, those that come closest achieve this due to interhemispheric warming differences rather than equatorial dynamics as observed. Addressing different origins of subsurface cooling in observations and simulations, and how they connect to SST, will lead to improved understanding of tropical Pacific SST changes to date and how they will evolve in the future.