Yichun He

This paper introduces growth-adaptive spring electronics and a spike processing pipeline to enable long-term, same-neuron mapping in developing rat brains, revealing neuron-specific developmental decorrelation.

This paper develops Deep-STARmap and Deep-RIBOmap for 3D spatial transcriptomics and translatomics in thick tissue blocks, enabling analysis of gene expression and immune interactions in human skin cancer.

This paper introduces BCI-Agent, an AI framework that classifies neuronal cell types from electrophysiological data without fine-tuning, validated in rodent and human datasets.

This paper introduces a stretchable microelectrode array that integrates into the developing brain, enabling long-term, stable neural activity tracking without disrupting development.

This paper introduces FlexiSoft, a flexible and robust electronic platform for brain-computer interfaces that enables long-term neural recording and curved implantation.

SpatialAgent integrates large language models with dynamic tool retrieval, multimodal interpretation, and claim verification to automate end-to-end spatial biology workflows from gene-panel design to tissue annotation and hypothesis generation.

The paper introduces STAgent, an AI tool that automates spatial transcriptomics analysis, revealing in vivo maturation of human stem cell-derived pancreatic cells.

SpikeAgent automates spike sorting with a multimodal LLM-based AI agent, achieving curation consistency equal to or exceeding human experts and drastically reducing expertise barriers.

This paper introduces AutoSort, a deep learning method for real-time decoding of neural activity from the same cells over months, showing improved efficiency and accuracy compared to existing methods.

The paper introduces FuseMap, a deep-learning framework for spatial transcriptomics that creates a comprehensive molecular spatial reference atlas of the mouse brain.

The paper creates a high-resolution spatial atlas of the mouse CNS by combining in situ sequencing with scRNA-seq data to map cell types and gene expression across the brain and spinal cord.

This paper introduces TEMPOmap, a method to profile subcellular RNA kinetics in single cells, revealing kinetic gene clusters with distinct functional roles.

This paper introduces in situ electro-sequencing to map gene expression and electrophysiology in 3D tissues, improving cell type analysis and developmental trajectory reconstruction.

This study uses machine learning to optimize organoid morphogenesis, revealing an excitable system driven by WNT and FGF signaling that governs human axial elongation.

This paper introduces UnitedNet, a multi-task deep learning model for analyzing single-cell multi-modality data, showing improved performance in joint group identification and cross-modal prediction.

ClusterMap is a framework for analyzing spatial gene expression data, clustering RNAs into subcellular structures and tissue regions across multiple dimensions.