Joseph O. Deasy

NetFlow constructs graph representations of biomedical data to reveal structured patterns and refine clinical subtypes in cancer datasets.

ORCO is an unsupervised network-based tool that computes Ollivier-Ricci curvature to analyze biological network robustness and identify therapeutic targets.

This study identifies high WEE1 expression as a prognostic marker in multiple myeloma, showing it is independently linked to poor survival and associated with dysregulated P53 pathways.

This study evaluates how proton therapy's tumor control effectiveness varies with fractionation schedules in early-stage NSCLC, finding higher RBE for short-term treatments.

This paper introduces a cloud-based framework for AI medical image analysis using pyCERR and Cancer Genomics Cloud, enabling end-to-end workflows without local software installation.

This study examines how CT scan reconstruction kernels and tube current settings affect radiomic features, finding that some features are robust while others vary significantly.

This study identifies gene modules associated with melanoma immunotherapy response using dynamic Ollivier-Ricci curvature analysis of gene expression data.

This paper introduces a network curvature framework to identify cancer subtypes from multi-omic data, showing significant survival differences across cancers.

STRATIS provides cloud-enabled software workflows for AI modeling in radiotherapy and imaging analysis, supporting multi-institutional collaboration and deployment.

This paper introduces a method based on the maximum entropy principle to model compositional data, demonstrating its effectiveness in capturing inter-species relationships and gene-gene interactions.

The paper develops an NLP/ML model to identify DILI-related literature using titles and abstracts, achieving high accuracy across multiple validation sets.

This study uses geometric network analysis of RNA-Seq data to cluster pediatric sarcoma tumors and identify therapeutic targets through unsupervised methods.

This paper develops deep learning models to automate segmentation of swallowing and chewing structures in H&N CT scans, achieving high accuracy and clinical utility.

The paper analyzes two single-cell datasets to study how p21 and p53 dynamics influence cell cycle arrest and mitosis frequency after radiation exposure.

The paper introduces vWCluster, a network-based clustering method for multi-omics breast cancer data using vector-valued optimal transport, identifying tumor subtypes with distinct survival rates and immune microenvironment profiles.

This paper presents a framework for deploying deep learning segmentation models for medical images across operating systems and languages, enabling clinical use and multi-institutional studies.

The paper introduces a periodicity scoring algorithm for p53 time series and shows its correlation with cell division events.

The paper presents an open-source library of deep-learning models for image segmentation and outcomes in oncology, enabling validation, automation, and integration into clinical workflows.

This study introduces a network-based method combining optimal mass transport and machine learning to predict drug sensitivity in cancer cell lines by clustering genomic and chemical features.

This paper introduces iWCluster, a network-based multiomics clustering method for breast cancer data, which recovers known subtypes and identifies hypoxia-related genes associated with poor prognosis.

The paper introduces a framework using optimal mass transport to construct a network reduction hierarchy for interactive exploration and community detection, applied to gene regulatory networks and validated on biological datasets.

This study introduces an optimal mass transport model to analyze contrast agent flow in head and neck cancer patients, showing comparable results to traditional methods.

This study develops CT-based texture analysis to quantify intra- and inter-site tumor heterogeneity in HGSOC and shows it correlates with progression-free survival.