Table 1.
Comparison with other tools.
Fig 1.
Based on 1D tomo-seq data along three mutually orthogonal axes and a mask data, tomoseqr performs RNA tomography using iterative proportional fitting and reconstructs a 3D expression pattern of each gene. A GUI called masker helps users to design and edit a mask data. Another GUI Image viewer visualizes the reconstructed gene expression patterns in 2D or 3D view.
Fig 2.
User interface of masker.
Fig 3.
User interface of Image viewer (2D view).
Fig 4.
User interface of Image viewer (3D view).
Fig 5.
Spatial expression patterns and reconstruction results for Gene1, Gene2, Gene3, Gene4, and Gene5.
Ground truth and reconstruction results are shown for (A) Gene1, (B) Gene2, (C) Gene3, (D) Gene4, and (E) Gene5. The dots are colored by gene expression levels, with white dots representing the sample shape (mask).
Table 2.
Evaluation of the reconstruction accuracy of tomoseq using simulated tomo-seq data.
For each simulating setting and spatial gene, 10 simulated tomo-seq data were generated. tomoseqr was applied to these data with the inter-section normalization option. Means and standard deviations of PCCs of the reconstruction results with the ground truth across 10 simulated tomo-seq data are shown. P-values were calculated by randomly shuffling the reconstructed results and calculating PCCs of the randomized results with the ground truth 1,000 times, following Kolmogorov-Smirnov tests (Materials and methods).
Table 3.
Computation time and amount of memory usage for reconstruction using tomoseqr.
Fig 6.
Results of reconstruction with zebrafish (Danio rerio, shield stage).
Reconstruction result for expression of (A) eve1 (ENSDARG00000056012), (B) bmp7a (ENSDARG00000018260), (C) ntla (ENSDARG00000009905), and (D) chd (ENSDARG00000006110). For each of the genes, 3D view, cross section perpendicular to the animal-vegetal axis (18 m × 50 sections), cross section perpendicular to the dorsal-ventral axis (18 m × 49 sections), and cross section perpendicular to the left-right axis (18 m × 56 sections) are shown from left to right. Parentheses after gene names indicate gene IDs. Colors represent the reconstructed expression levels.
Fig 7.
Reconstruction results using planarian tomo-seq data.
Reconstruction result for expression of (A) piwiA (DjGI005146_001), (B) opsin (DjGI008464_001), (C) Djf-1 (DjGI006376_001), and (D) DjNp19 (DjGI017609_001). Parentheses after gene names indicate gene IDs. Colors represent the reconstructed expression levels.AP means anterior-posterior axis (80 μm × 78 sections), VD means ventral-dorsal axis,(20 μm × 28 sections) and LR means left-right axis (20 μm × 51 sections).
Fig 8.
Systematic exploration of genes correlated with the reconstructed opsin expression pattern in planarian tomo-seq data.
(A) A histogram showing the Pearson correlation coefficients (PCCs) with the reconstructed opsin expression pattern for 18,768 expressed genes. The vertical lines indicate the top 4 genes with the highest PCCs. (B-E) Reconstruction results for the expression of (B) Arrb1 (DjGI008623_001), (C) Rnf13 (DjGI009330_001), (D) Cpne9 (DjGI006974_001), and (E) Tnnc1 (DjGI015938_001). The PCCs of these gene expression patterns with the opsin gene (Fig 7B) are shown. (F) A heatmap showing the expression of human homologs in human eye single-cell types using single-cell RNA sequencing data from the Human Protein Atlas. The gene expression levels, in normalized transcript per million (nTPM), were transformed into Z-scores.
Fig 9.
Systematic exploration of genes with spatial autocorrelation in the reconstructed expression patterns in planarian tomo-seq data.
(A) A histogram showing the Moran’s I values of the reconstructed expression patterns for 214 genes encoding planarian homologs of mouse transcription factors. (B-F) Reconstruction results for the expression of (B) Creb3l1 (DjGI019398_001), (C) Kmt2d (DjGI001455_001), (D) Smad1 (DjGI002591_001), (E) Mzf1 (DjGI000932_002), and (F) Foxa2 (DjGI007454_001). The corresponding Moran’s I values are shown.