autoencoder  (Nissen)

Journal: Journal of Translational Medicine

Article Title: HiSTaR: identifying spatial domains with hierarchical spatial transcriptomics variational autoencoder

doi: 10.1186/s12967-025-07404-3

Figure Lengend Snippet: Overview of HiSTaR framework. Framework comprises three components: data process, HiSTaR (encoder and decoder), and downstream analysis. In data process, the gene expression matrix is randomly masked, and the adjacency matrix is constructed. The encoder includes a fully connected network and two-level HiSTaR blocks to capture multilevel latent representations and the decoder reconstructs the expression matrix and the adjacency matrix. Latent representations can be used to perform downstream analyses, including spatial domain identification and batch-effects correction

Article Snippet: In this paper, we propose a Hierarchical Spatial Transcriptomics variational autoencoder (HiSTaR) that employs multiple HiSTaR blocks to capture multi-level latent features from spots.

Techniques: Gene Expression, Construct, Expressing

Journal: Journal of Translational Medicine

Article Title: HiSTaR: identifying spatial domains with hierarchical spatial transcriptomics variational autoencoder

doi: 10.1186/s12967-025-07404-3

Figure Lengend Snippet: HiSTaR is evaluated in human DLPFC dataset. ( a ) Ground truth of spatial domains in slice #151673 from the human DLPFC dataset. ( b ) Spatial domain identification results on slice #151673 from the human DLPFC dataset, generated by Leiden, Louvain, ST-SCSR, sedr, STAGATE, STMGraph, DeepST, and HiSTaR. ( c ) Quantitative comparison of spatial domain identification across all 12 slices of the human DLPFC dataset, evaluated using ARI, NMI, and FMS metrics. ( d ) Umap plots and PAGA visualizations of the latent space representations learned by STAGATE, STMGraph, and HiSTaR on slice #151673 from the human DLPFC dataset

Article Snippet: In this paper, we propose a Hierarchical Spatial Transcriptomics variational autoencoder (HiSTaR) that employs multiple HiSTaR blocks to capture multi-level latent features from spots.

Techniques: Generated, Comparison

Journal: Journal of Translational Medicine

Article Title: HiSTaR: identifying spatial domains with hierarchical spatial transcriptomics variational autoencoder

doi: 10.1186/s12967-025-07404-3

Figure Lengend Snippet: HiSTaR is evaluated on human breast cancer dataset. ( a ) Ground truth annotations of spatial domains in the human breast cancer dataset. ( b ) Spatial domain identification results on the human breast cancer dataset, generated by STAGATE, STMGraph, and HiSTaR. ( c ) Dot plot of the top three differentially expressed genes (DEGs) in each spatial domain. Circle color indicates gene expression level, while circle size reflects the proportion of spots expressing the gene. ( d ) Representative spatial expression patterns of MS4A1 (left) and KRT14 (right)

Article Snippet: In this paper, we propose a Hierarchical Spatial Transcriptomics variational autoencoder (HiSTaR) that employs multiple HiSTaR blocks to capture multi-level latent features from spots.

Techniques: Generated, Gene Expression, Expressing

Journal: Journal of Translational Medicine

Article Title: HiSTaR: identifying spatial domains with hierarchical spatial transcriptomics variational autoencoder

doi: 10.1186/s12967-025-07404-3

Figure Lengend Snippet: HiSTaR is evaluated on multiple datasets from diverse platforms. ( a ) Reference atlas from the Allen Mouse Brain atlas. ( b ) H&E-stained image of the mouse brain tissue section. ( c ) Spatial domain identification results and UMAP plots of the mouse brain dataset generated by STAGATE, STMGraph, and HiSTaR, based on the 10x genomics visium platform. ( d ) DAPI-stained image of the mouse olfactory bulb. ( e ) Spatial domain identification results and UMAP plots of the mouse olfactory bulb dataset generated by STAGATE, STMGraph, and HiSTaR, based on the stereo-seq platform. ( f ) Reference atlas from the Allen Mouse olfactory bulb atlas. ( g ) Spatial domain identification results of the mouse olfactory bulb dataset generated by STAGATE, STMGraph, and HiSTaR, based on the silde-seqV2 platform. ( h ) Spatial domains identified by HiSTaR (upper) and the corresponding expression patterns of a specific biomarker genes (lower). ( i ) Ground truth of mouse visual cortex dataset. ( j ) Spatial domain identified by STAGATE, STMGraph, and HiSTaR, based on the STARmap platform

Article Snippet: In this paper, we propose a Hierarchical Spatial Transcriptomics variational autoencoder (HiSTaR) that employs multiple HiSTaR blocks to capture multi-level latent features from spots.

Techniques: Staining, Generated, Expressing, Biomarker Discovery

Journal: Journal of Translational Medicine

Article Title: HiSTaR: identifying spatial domains with hierarchical spatial transcriptomics variational autoencoder

doi: 10.1186/s12967-025-07404-3

Figure Lengend Snippet: HiSTaR corrects batch-effects on multiple slices. ( a ) Vertical alignment of spatial Sect. and Sect. from the mouse breast cancer dataset. ( b ) UMAP plot of Sect. and Sect. . ( c ) Boxplot of iLISI scores for STAGATE, STMGraph, and HiSTaR. ( d ) Clustering results (left) and spatial domain identification results (right) after batch-effects correction by STAGATE, STMGraph, and HiSTaR. ( e ) Vertical alignment of spatial Sect. and Sect. from the mouse brain dataset. ( f ) UMAP plot of Sect. and Sect. . ( g ) Boxplot of iLISI scores for STAGATE, STMGraph, and HiSTaR. ( h ) Clustering results (left) and spatial domain identification results (right) after batch-effects correction by STAGATE, STMGraph, and HiSTaR. ( i ) Alignment of anterior and posterior sections from the mouse brain dataset. ( j ) Reference annotations from the Allen mouse brain Atlas. ( k ) H&E-stained images of aligned anterior and posterior, with PkH, MolH, and IGrH. ( l ) Spatial domain after batch-effects correction by STAGATE, STMGraph, and HiSTaR

Article Snippet: In this paper, we propose a Hierarchical Spatial Transcriptomics variational autoencoder (HiSTaR) that employs multiple HiSTaR blocks to capture multi-level latent features from spots.

Techniques: Staining

Journal: MethodsX

Article Title: Analysis of the impact of irradiance and temperature on photovoltaic production: A statistical and machine learning approach

doi: 10.1016/j.mex.2025.103716

Figure Lengend Snippet: Autoencoder architecture.

Article Snippet: Second, we introduce a machine learning-based approach specifically using an autoencoder architecture to model the complex, nonlinear dynamics inherent in PV systems [ ].

Techniques:

Journal: MethodsX

Article Title: Analysis of the impact of irradiance and temperature on photovoltaic production: A statistical and machine learning approach

doi: 10.1016/j.mex.2025.103716

Figure Lengend Snippet: Response surface of PV production predicted by the autoencoder.

Article Snippet: Second, we introduce a machine learning-based approach specifically using an autoencoder architecture to model the complex, nonlinear dynamics inherent in PV systems [ ].

Techniques:

Journal: MethodsX

Article Title: Analysis of the impact of irradiance and temperature on photovoltaic production: A statistical and machine learning approach

doi: 10.1016/j.mex.2025.103716

Figure Lengend Snippet: Seasonal performance comparison of autoencoder and regression for photovoltaic production.

Article Snippet: Second, we introduce a machine learning-based approach specifically using an autoencoder architecture to model the complex, nonlinear dynamics inherent in PV systems [ ].

Techniques: Comparison