Research

Create a single cell atlas of the human gut across age, geography, and ethnicity to better understand
the healthy gut of developing children worldwide (Mapping the Early Childhood Gut Across Ancestry, Geography and Environment – Chan Zuckerberg Initiative)

Use innovative data science techniques to identify deep features of EE, correlate these features to disease pathogenesis, and discern disease specific features which may inform future interventions
(RePORT ⟩ RePORTER (nih.gov)

Generate a computational model of the human ileum using histology and ‘omics data from a prospective pediatric CD cohort to identify metabolic pathways and biomarkers associated with Crohn’s disease. These pathways will be validated and targeted in patient derived enteroids to determine whether these pathways contribute to Crohn’s pathogenesis (RePORT ⟩ RePORTER (nih.gov)

Build machine learning models capable of predicting future outcomes of Crohn’s disease (anti-TNF therapy resistance, structural complications from disease progression) from initial endoscopy tissue acquired from large archival and prospective pediatric Crohn’s cohorts. In parallel, cutting-edge single cell techniques will be used to create a genomic profile of Crohn’s disease at a single-cell resolution to better inform our models and reveal underlying genetic changes associated with disease morphology.

• Computational Characterization of Environmental Enteropathy
  • Build & deconstruct a Deep Learning Net to enhance detection of morphological features of environmental Enteropathy (EE) versus celiac and healthy small intestinal tissue. We will also correlate distinguishing morphological EE features with clinical phenotypes, measures of gut barrier and absorption, microbiome, and bile acid deconjugation. Funding: NIH (2019), iTHRIV-UVA (2018), Engineering in Medicine-UVA (2017)

• Use of machine learning image analysis and tissue transcriptomics to define clinically actionable celiac disease sub-types
  • This project aims to address the pressing questions for the role of enteroendocrine cells in CD severity and to what extent are they linked to the patients who go on to develop autoimmune endocrinopathies such as diabetes or hypothyroidism. Funding: iTHRIV Pilot Translational and Clinical Studies (2021)

• Modeling Crohn’s disease using Machine Learning: Image Analysis and Multiomics
  • Construction of biologically informative and clinically useful diagnostic and prognostic models that capture complex Crohn’s disease phenotypes – using ileal pathology images and further correlations with clinical, ‘omic, and molecular metadata along with leveraging metabolic network modeling. Funding: Crohn’s & Colitis Foundation (2020)

• Computational Prediction of Crohn’s Disease Clinical Phenotypes via Magnetic Resonance Imaging.
  • Investigating distinguishing patterns of Crohn’s disease clinical phenotypes using 3D reconstruction of radiology imaging data (Magnetic Resonance Enterography/Imaging).

• Understanding Ulcerative Colitis via Deep Learning and Image Segmentation
  • Evaluation of Ulcerative Colitis tissue eosinophilia and its correlation with steroid or biologic therapy.

• Characterizing Eosinophilic Esophagitis with Deep Learning and Image Segmentation
  • Evaluation of Eosinophilic Esophagitis phenotypes using biopsy images via the development of deep learning architectures.

• Mapping the Early Childhood Gut Across Ancestry, Geography, and Environment
  • This project will map early gut development across populations with diverse ancestry and geography, at single-cell resolution, and with linked contextual data on tissue morphology, genetic background, social determinants of health, and environmental exposures in pediatric patients aged 0-5 years. Funding: CZI (2021)

• Video Capsule Endoscopy based Gastrointestinal Tissue Video Analysis using Deep Learning
  • Video analysis platform for automated extraction of quantitative morphologic phenotypes from subjects with healthy and diseased gastrointestinal tissue. Funding: Engineering-in-Medicine, UVA (2020)