Harnessing heterogeneity in space with statistically guided meta-learning

Spatial data are ubiquitous, massively collected, and widely used to support critical decision-making in many societal domains, including public health (e.g., COVID-19 pandemic control), agricultural crop monitoring, transportation, etc. While recent advances in machine learning and deep learning offer new promising ways to mine such rich datasets (e.g., satellite imagery, COVID statistics), spatial heterogeneity—an intrinsic characteristic embedded in spatial data—poses a major challenge as data distributions or generative processes often vary across space at different scales, with their spatial extents unknown. Recent studies (e.g., SVANN, spatial ensemble) targeting this difficult problem either require a known space-partitioning as the input, or can only support very limited number of partitions or classes (e.g., two) due to the decrease in training data size and the complexity of analysis. To address these limitations, we propose a model-agnostic framework to automatically transform a deep learning model into a spatial-heterogeneity-aware architecture, where the learning of arbitrary space partitionings is guided by a learning-engaged generalization of multivariate scan statistic and parameters are shared based on spatial relationships. Moreover, we propose a spatial moderator to generalize learned space partitionings to new test regions. Finally, we extend the framework by integrating meta-learning-based training strategies into both spatial transformation and moderation to enhance knowledge sharing and adaptation among different processes. Experiment results on real-world datasets show that the framework can effectively capture flexibly shaped heterogeneous footprints and substantially improve prediction performances.

Harnessing heterogeneity in space with statistically guided meta-learning.

Spatial data are ubiquitous, massively collected, and widely used to support critical decision-making in many societal domains, including public health (e.g., COVID-19 pandemic control), agricultural crop monitoring, transportation, etc. While recent advances in machine learning and deep learning offer new promising ways to mine such rich datasets (e.g., satellite imagery, COVID statistics), spatial heterogeneity-an intrinsic characteristic embedded in spatial data-poses a major challenge as data distributions or generative processes often vary across space at different scales, with their spatial extents unknown. Recent studies (e.g., SVANN, spatial ensemble) targeting this difficult problem either require a known space-partitioning as the input, or can only support very limited number of partitions or classes (e.g., two) due to the decrease in training data size and the complexity of analysis. To address these limitations, we propose a model-agnostic framework to automatically transform a deep learning model into a spatial-heterogeneity-aware architecture, where the learning of arbitrary space partitionings is guided by a learning-engaged generalization of multivariate scan statistic and parameters are shared based on spatial relationships. Moreover, we propose a spatial moderator to generalize learned space partitionings to new test regions. Finally, we extend the framework by integrating meta-learning-based training strategies into both spatial transformation and moderation to enhance knowledge sharing and adaptation among different processes. Experiment results on real-world datasets show that the framework can effectively capture flexibly shaped heterogeneous footprints and substantially improve prediction performances.

Alpha-Soluble NSF Attachment Protein Prevents the Cleavage of the SARS-CoV-2 Spike Protein by Functioning as an Interferon-Upregulated Furin Inhibitor.

Loss of the furin cleavage motif in the SARS-CoV-2 spike protein reduces the virulence and transmission of SARS-CoV-2, suggesting that furin is an attractive antiviral drug target. However, lack of understanding of the regulation of furin activity has largely limited the development of furin-based therapeutic strategies. Here, we find that alpha-soluble NSF attachment protein (α-SNAP), an indispensable component of vesicle trafficking machinery, inhibits the cleavage of SARS-CoV-2 spike protein and other furin-dependent virus glycoproteins. SARS-CoV-2 infection increases the expression of α-SNAP, and overexpression of α-SNAP reduces SARS-CoV-2 infection in cells. We further reveal that α-SNAP is an interferon-upregulated furin inhibitor that inhibits furin function by interacting with its P domain. Our study demonstrates that α-SNAP, in addition to its role in vesicle trafficking, plays an important role in the host defense against furin-dependent virus infection and therefore could be a target for the development of therapeutic options for COVID-19. IMPORTANCE Some key mutations of SARS-CoV-2 spike protein, such as D614G and P681R mutations, increase the transmission or pathogenicity by enhancing the cleavage efficacy of spike protein by furin. Loss of the furin cleavage motif of SARS-CoV-2 spike protein reduces the virulence and transmission, suggesting that furin is an attractive antiviral drug target. However, lack of understanding of the regulation of furin activity has largely limited the development of furin-based therapeutic strategies. Here, we found that in addition to its canonical role in vesicle trafficking, alpha-soluble NSF attachment protein (α-SNAP) plays an important role in the host defense against furin-dependent virus infection. we identified that α-SNAP is a novel interferon-upregulated furin inhibitor and inhibits the cleavage of SARS-CoV-2 spike protein and other furin-dependent virus glycoproteins by interacting with P domain of furin. Our study demonstrates that α-SNAP could be a target for the development of therapeutic options for COVID-19.

SARS-CoV-2 uses metabotropic glutamate receptor subtype 2 as an internalization factor to infect cells.

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) uses angiotensin-converting enzyme 2 (ACE2) as a binding receptor to enter cells via clathrin-mediated endocytosis (CME). However, receptors involved in other steps of SARS-CoV-2 infection remain largely unknown. Here, we found that metabotropic glutamate receptor subtype 2 (mGluR2) is an internalization factor for SARS-CoV-2. Our results show that mGluR2 directly interacts with the SARS-CoV-2 spike protein and that knockdown of mGluR2 decreases internalization of SARS-CoV-2 but not cell binding. Further, mGluR2 is uncovered to cooperate with ACE2 to facilitate SARS-CoV-2 internalization through CME and mGluR2 knockout in mice abolished SARS-CoV-2 infection in the nasal turbinates and significantly reduced viral infection in the lungs. Notably, mGluR2 is also important for SARS-CoV spike protein- and Middle East respiratory syndrome coronavirus spike protein-mediated internalization. Thus, our study identifies a novel internalization factor used by SARS-CoV-2 and opens a new door for antiviral development against coronavirus infection.