Daifeng Wang, PhD – Slide of the Week
Title: SegJointGene: joint cell segmentation and spatial gene prioritization by information entropy guided convolutional neural networks
Legend: Overview of the SegJointGene framework. The framework integrates spatial transcriptomics to perform joint spatial gene prioritization and cell-segmentation refinement. Input preparation: Spatial transcriptomics data (mRNA spot coordinates, single-cell gene expression, and cell-nuclei images) are processed into two inputs. Input 1: X is the stack of gene-density maps derived from mRNA spot coordinates, where each map Xj represents the spatial distribution of gene j. Input 2: A is the initial cell-segmentation map derived from cell-nuclei staining and cell-type annotation, assigning a cell-type label k to each cell region. SegJointGene: A convolutional neural network (CNN) inputs the gene-density maps X and the initial cell-segmentation map A as input. The network uses feature extraction and information-entropy calculation to iteratively optimize the segmentation parameters. The framework generates two outputs: Output 1: Sjk, the Gene Importance Score quantifying the spatial relevance of gene j for identifying cell type k; and Output 2: A*, the final cell-segmentation map with refined cell-type assignments k.
Citation: Haotian Ma, Daifeng Wang, SegJointGene: joint cell segmentation and spatial gene prioritization by information entropy guided convolutional neural networks, Bioinformatics, Volume 42, Issue 7, July 2026, btag447, https://doi.org/10.1093/bioinformatics/btag447
Abstract: Spatial sequencing technologies enable the single-cell-level study of molecular organization in tissues. Revealing such spatial patterns relies on accurate cell segmentation. In complex tissues with dense cell packing, segmentation based solely on nuclear staining is insufficient for accurate cell boundary detection. This limitation arises because accurate segmentation necessitates the delineation of cell morphology, which is driven by molecular activities such as cytoskeletal dynamics, cell-cell adhesion, and intercellular signaling. Thus, integrating molecular information, including gene or protein expression, has the potential to improve segmentation, but remains computationally challenging.
Investigator: Daifeng Wang, PhD
About the Lab: The Wang lab develops machine learning and artificial intelligence (ML/AI) approaches and bioinformatics tools for understanding the cellular and molecular mechanisms from genotypes to phenotypes in complex brains (such as single-cell multimodal learning as above). Their applications focus on functional genomics, gene regulation, and neural circuits, particularly for brain development, intelligence, and brain diseases.