Anqi Zou
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Anqi Zou has a total of 20 co-authors. Here are the top co-authors by number of shared publications:
- Chao Zhang
Department of Medicine, Boston University School of Medicine, Boston, USA
3 shared publications - Deborah Lang
Department of Dermatology, Boston University, Boston, MA 02118, USA
1 shared publication Elain Huang
Section of Computational Biomedicine Section, Department of Medicine, Boston University Chobanian & Avedisian School of Medicine, Boston, MA, USA
1 shared publicationEmily Oei
Nephrology Section, Department of Medicine, Boston University Chobanian & Avedisian School of Medicine, Boston Medical Center, Boston, MA, USA
1 shared publicationFlorencia Rojas-Miguez
Department of Pathology and Laboratory Medicine, Boston University Chobanian & Avedisian School of Medicine, Boston Medical Center, Boston, MA, USA
1 shared publicationHongying Yang
Global Pathology and Centers for Therapeutic Innovation, Pfizer, Inc., Groton, CT, and Cambridge, MA, USA
1 shared publication
Latest Publications
- Open Access
RNA-guided contrastive learning enhances patient-level prediction from histology
Predicting molecular receptor status, including estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2), directly from routine hematoxylin and eosin (H&E)-stained whole-slide images (WSIs) could reduce reliance on costly immunohistochemistry, but image-only models have no access to the transcriptomic programs that define these subtypes. Here we show that coupling a whole-slide vision transformer (GigaPath) with bulk RNA-sequencing profiles through a training-time-only contrastive objective improves receptor-status classification from WSIs alone, without requiring any RNA-seq data at inference. Rather than training a dedicated bulk-transcriptomics encoder, we purpose the gene-sentence encoding and pretrained text encoder from BioBERT 1 and apply them to bulk RNA-seq profiles from The Cancer Genome Atlas breast cancer cohort (TCGA-BRCA). Under five-fold cross-validation, this RNA-guided pre-training consistently improved AUROC for ER, PR, and HER2 relative to a contrastive-pre-training-free baseline and reproducibly introduced receptor-relevant structure into the learned slide representation. However, evaluation on independent external cohorts revealed a striking, receptor-dependent divergence: gains for ER and PR, both governed by broad, multi-gene luminal transcriptional programs, generalized robustly, whereas the apparent AUROC improvement for HER2, driven by focal ERBB2 amplification rather than a coordinated transcriptional signature, masked a severe loss of classification sensitivity at the operating threshold. These findings establish training-time transcriptomic supervision as a practical route to enhancing histology-only molecular subtyping, while showing that gene-selection strategy, and not model architecture alone, determines which molecular alterations such supervision can teach a vision model to recognize.
Open AccessRNA-guided contrastive learning enhances patient-level prediction from histology
Predicting molecular receptor status, including estrogen receptor (ER), progesterone receptor (PR), and human epidermal growth factor receptor 2 (HER2), directly from routine hematoxylin and eosin (H&E)-stained whole-slide images (WSIs) could reduce reliance on costly immunohistochemistry, but image-only models have no access to the transcriptomic programs that define these subtypes. Here we show that coupling a whole-slide vision transformer (GigaPath) with bulk RNA-sequencing profiles through a training-time-only contrastive objective improves receptor-status classification from WSIs alone, without requiring any RNA-seq data at inference. Rather than training a dedicated bulk-transcriptomics encoder, we purpose the gene-sentence encoding and pretrained text encoder from BioBERT 1 and apply them to bulk RNA-seq profiles from The Cancer Genome Atlas breast cancer cohort (TCGA-BRCA). Under five-fold cross-validation, this RNA-guided pre-training consistently improved AUROC for ER, PR, and HER2 relative to a contrastive-pre-training-free baseline and reproducibly introduced receptor-relevant structure into the learned slide representation. However, evaluation on independent external cohorts revealed a striking, receptor-dependent divergence: gains for ER and PR, both governed by broad, multi-gene luminal transcriptional programs, generalized robustly, whereas the apparent AUROC improvement for HER2, driven by focal ERBB2 amplification rather than a coordinated transcriptional signature, masked a severe loss of classification sensitivity at the operating threshold. These findings establish training-time transcriptomic supervision as a practical route to enhancing histology-only molecular subtyping, while showing that gene-selection strategy, and not model architecture alone, determines which molecular alterations such supervision can teach a vision model to recognize.
- Open Access
Deep Learning Enables Automated Segmentation and Quantification of Ultrastructure from Transmission Electron Microscopy Images
Transmission electron microscopy (TEM) has become an essential technique for observing subcellular ultrastructure, and is widely used in both clinical diagnosis and biomedical research. However, analysis of TEM data remains extremely labor-intensive and often inconsistent across operators due to the lack of dedicated computational methods. Here, we present TEAMKidney, a deep learning framework for accurate and scalable measurement of ultrastructures in TEM images across species, magnifications, and instrument platforms. We collected 12,991 TEM images from patients with multiple kidney diseases and from different animal models. By combining a self-training-based semantic segmentation stage with a TEM-tailored panoptic segmentation model, we address two major challenges in TEM data analysis: the lack of accurately labeled training data and the difficulty of achieving high segmentation accuracy for complex ultrastructure. Application of TEAMKidney to both human and animal images successfully reveals disease-associated changes in two critical glomerular ultrastructures: the glomerular basement membrane and podocyte foot processes. In addition to significantly outperforming existing tools, TEAMKidney shows close agreement with pathological expert measurements used in clinical assessment protocols. By reducing dependence on manual tracing while preserving expert-level accuracy, TEAMKidney demonstrates that deep learning can substantially reduce the burden of image analysis in both clinical pathology and biomedical research settings.
Open AccessDeep Learning Enables Automated Segmentation and Quantification of Ultrastructure from Transmission Electron Microscopy Images
Transmission electron microscopy (TEM) has become an essential technique for observing subcellular ultrastructure, and is widely used in both clinical diagnosis and biomedical research. However, analysis of TEM data remains extremely labor-intensive and often inconsistent across operators due to the lack of dedicated computational methods. Here, we present TEAMKidney, a deep learning framework for accurate and scalable measurement of ultrastructures in TEM images across species, magnifications, and instrument platforms. We collected 12,991 TEM images from patients with multiple kidney diseases and from different animal models. By combining a self-training-based semantic segmentation stage with a TEM-tailored panoptic segmentation model, we address two major challenges in TEM data analysis: the lack of accurately labeled training data and the difficulty of achieving high segmentation accuracy for complex ultrastructure. Application of TEAMKidney to both human and animal images successfully reveals disease-associated changes in two critical glomerular ultrastructures: the glomerular basement membrane and podocyte foot processes. In addition to significantly outperforming existing tools, TEAMKidney shows close agreement with pathological expert measurements used in clinical assessment protocols. By reducing dependence on manual tracing while preserving expert-level accuracy, TEAMKidney demonstrates that deep learning can substantially reduce the burden of image analysis in both clinical pathology and biomedical research settings.
- Open Access
- 24 October 2025
- Published by Springer Nature in BMC Bioinformatics
- Vol. 26 (1) , 1-11
VaMiAnalyzer: an open source, Python-based application for analysis of 3D in vitro vasculogenic mimicry assays
Vasculogenic mimicry (VM) is the phenomenon whereby non-vascular tumor cells develop vascular-like structures. VM is linked to more aggressive tumor phenotypes including higher rates of metastasis and invasion and is potentially resistant to anti-angiogenic cancer therapies. VM is investigated in vitro using 3D assays with microscopy images capturing the resulting VM structures, including loops, branch points, and tubes. The standard method to quantify endpoint data is to count various structural features manually, which is time-consuming and open to bias. At present, no software solutions have been developed to specifically address the analysis and quantification of VM structures. To address this limitation, we developed an open source, Python-based application, VaMiAnalyzer, allowing straightforward quantification of several VM structural features. The application follows a two-step approach that optionally corrects and enhances the raw input images and then analyzes and quantifies the VM features. VaMiAnalyzer is stand-alone software that allows automated measurement of VM structural features from phase-contrast microscopy images. It produces results that are strongly consistent with manual counts but in a significantly shorter time, allowing quick, non-biased analysis of VM from microscopy images.
- 24 October 2025
- Published by Springer Nature in BMC Bioinformatics
- Vol. 26 (1) , 1-11
Open AccessVaMiAnalyzer: an open source, Python-based application for analysis of 3D in vitro vasculogenic mimicry assays
Vasculogenic mimicry (VM) is the phenomenon whereby non-vascular tumor cells develop vascular-like structures. VM is linked to more aggressive tumor phenotypes including higher rates of metastasis and invasion and is potentially resistant to anti-angiogenic cancer therapies. VM is investigated in vitro using 3D assays with microscopy images capturing the resulting VM structures, including loops, branch points, and tubes. The standard method to quantify endpoint data is to count various structural features manually, which is time-consuming and open to bias. At present, no software solutions have been developed to specifically address the analysis and quantification of VM structures. To address this limitation, we developed an open source, Python-based application, VaMiAnalyzer, allowing straightforward quantification of several VM structural features. The application follows a two-step approach that optionally corrects and enhances the raw input images and then analyzes and quantifies the VM features. VaMiAnalyzer is stand-alone software that allows automated measurement of VM structural features from phase-contrast microscopy images. It produces results that are strongly consistent with manual counts but in a significantly shorter time, allowing quick, non-biased analysis of VM from microscopy images.