About

Ryan J. Urbanowicz, Ph.D., is an Adjunct Assistant Professor of Biostatistics and Epidemiology at the University of Pennsylvania's Perelman School of Medicine. His research focuses on developing, evaluating, and applying novel computational, statistical, and visualization methods to facilitate classification and data mining in biomedical research, particularly in the context of complex, noisy data related to human diseases. His primary interests include the adaptation of learning classifier system algorithms to detect, model, and characterize epistatic and heterogeneous associations within SNP association studies, as well as creating strategies that allow data to speak for itself without making many assumptions. Dr. Urbanowicz's work has led to the development of ExSTraCS, an extended supervised tracking and classifying system, and the open-source GAMETES software package for complex disease model and data simulation. His research also involves tackling issues related to big data by developing new feature selection approaches such as ReliefF, SURF, and MultiSURF, which serve as critical preprocessing steps for feature selection and knowledge discovery. His interdisciplinary approach intersects genetics, genomics, biostatistics, epidemiology, machine learning, and computer science, emphasizing a quantitative biomedical research strategy that embraces the complexity of relationships between predictive factors and disease endpoints.

Research topics

• Political Science
• Medicine
• Medical emergency

Selected publications

• Abstract 2545: Genome-wide analysis of ctDNA-derived DNA palindromes for early detection of breast cancer

  Cancer Research · 2026-04-03

  article

  Abstract Background and Objective: Circulating tumor DNA (ctDNA) in the blood provides valuable information about all aspects of patient care, including real-time tumor burden and therapeutic targets. While ctDNA tests currently focus on detecting small mutations and abnormal DNA methylations in targetedgenomic regions, structural variants (SVs) are also common in tumors and could serve as important cancer biomarkers. Despite the growing use of liquid biopsy in cancer detection and management, an effective agnostic SV detection method remains a missing piece. To address this, we target DNA palindromes, a chromosomal structural abnormality also known as fold-back inversions and inverted repeats. We explored the potential of an approach called GAPF-Seq (Genome-wide Analysis of Palindrome Formation: GAPF with NGS) for ctDNA detection. GAPF-seq enriches palindromic DNA from very small amounts of genomic DNAthrough intramolecular annealing (Tanaka et al., Nat Genet 2005). Next-generation sequencing (NGS) analysis of the enriched DNA would enable us to identify SVs across the genome. Methods: Breast tumor DNA was processed by the successive denaturation and renaturation. DNA from palindromes would form double-stranded DNA by intramolecular annealing, while normal and nonpalindromic DNA would remain single-stranded, which would be eliminated by S1 nuclease. Tumor-derived DNA palindromes were amplified by PCR, sequenced by NGS, and subjected to bioinformatic analysis, including ROC validation and chromosomal distribution profiling. Results: (1) ROC analysis demonstrated high diagnostic accuracy. Using the top 1000 high coverage bins(HCBs) among the 1 million 1-kb bins in the genome, we found that the AUC value for tumor DNA calling was 0.9885, with sensitivity 92.3% and specificity 97.4%, enabling robust distinction between cancer and normal samples, including Stage I tumors. Comparable performance was achieved between 30 ng and 100 ng of DNA.(2) As GAPF-Seq is based on Structural variant detection genome-wide, it allows comprehensive genome-wide screening beyond single-gene mutation or a subset of CpG methylation analyses. Chromosomal mapping revealed uniform distribution of top 1000 HCBs in normal samples, while tumor samples exhibited chromosome-specific enrichment of HCBs. Subtype-specific analysis showed accumulation around the CCND1 gene on chromosome 11 in Luminal type, and enrichment near the ERBB2 gene on chromosome 17 in HER2 type. Conclusion: GAPF-Seq has the potential to enable accurate detection of early breast cancer even from minimal DNA input. Its agnostic, genome-wide profiling capability provides an additional benefit. Potential clinical applications include cancer screening and monitoring minimal residual disease. Citation Format: Fumie Igari, Hisashi Tanaka, Tamami Hyodo, Yuko Ishikawa, Tomoyuki Fujita, Michael Murata, Ryan Urbanowicz, Armando Giuliano. Genome-wide analysis of ctDNA-derived DNA palindromes for early detection of breast cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 2545.

  PublisherDOI

• PRIMARY-AI: outcomes-based standards to safeguard primary care in the AI era

  Nature Medicine · 2026-02-11 · 1 citations

  article
  PublisherDOI

• Stepwise Models Identify Risk of Kidney Allograft Failure Associated with HLA Amino Acid Mismatches Among Adult FSGS Recipients

  American Journal of Transplantation · 2025-08-01

  articleOpen access
  PublisherDOI

• Automated Machine Learning Tools for Data Science, Modeling, and Algorithm Benchmarking

  Proceedings of the Genetic and Evolutionary Computation Conference Companion · 2025-07-14

  article1st authorCorresponding
  PublisherDOI

• HLA Class I Amino Acid Mismatches That Distinguish Divergent HLA-A Allele Lineages Confer Incremental Risk of Kidney Allograft Failure Independent of Class II Mismatches

  American Journal of Transplantation · 2025-08-01

  article
  PublisherDOI

• A Novel Machine Learning Algorithm Discovers Mismatched HLA Class I Amino Acid Positions That Confer Incremental Kidney Graft Failure Risk Independent of HLA Class II

  American Journal of Transplantation · 2025-08-01

  articleSenior author
  PublisherDOI

• Rule-based Machine Learning: Separating Rule and Rule-Set Pareto-Optimization for Interpretable Noise-Agnostic Modeling

  Proceedings of the Genetic and Evolutionary Computation Conference · 2025-07-08

  articleOpen accessSenior author

  Rule-based machine learning (RBML) algorithms, e.g. learning classifier systems (LCSs), can capture complex relationships while yielding more interpretable models than most other machine learning algorithms. Traditional LCSs rely on a single fitness function for both rule and/or rule-set optimization. However, ideal rule vs. rule-set discovery often requires distinct and multiple objectives. Recently, hybrid-LCSs were proposed that explicitly separated the task of rule vs. rule-set discovery but relied on distinct single-objective or weighted multi-objective fitness functions. This study introduces a newly developed Heuristic Evolutionary Rule Optimization System (HEROS) that combines previous LCS innovations aimed at tackling noisy, larger-scale, classification tasks, while adopting separation of rule vs. rule-set evolution. Uniquely, HEROS employs a custom Pareto-front-based multi-objective fitness function (for rule discovery) and NSGA-II-style multi-objective optimization (for rule-set discovery) to solve both clean and noisy-signal classification problems agnostically. Rule discovery is driven by rule-accuracy and instance coverage objectives, while rule-set discovery is driven by prediction accuracy and rule-set size objectives. Using diverse simulated benchmark datasets, i.e. noisy (GAMETES) and clean (MUX), we demonstrate proof-of-principle that HEROS can directly discover accurate, highly-compact, interpretable, and ideal solutions when compared to the established 'ExSTraCS' RBML algorithm, without objective weightings or adjusting hyperparameters.

  PublisherDOI

• 0627 Social Risk Factors, Sleep Apnea and Cardiovascular Risk: Clinical Predictors in Community Health Centers

  SLEEP · 2025-05-01

  articleOpen access

  Abstract Introduction Racial, ethnic and socioeconomically disadvantaged minorities are more likely to experience insufficient sleep, sleep disorders, and negative cardiovascular (CV) outcomes. However, pathways linking health disparities to sleep disturbances and CV outcomes are largely underexplored. We leveraged electronic health record (EHR) data from the Accelerating Data Value Across a National Community Health Center Network (ADVANCE) Clinical Research Network (CRN) to identify social risk factor clusters, assess their association with obstructive sleep apnea (OSA), and determine relevant clinical predictors of cardiovascular (CV) outcomes among those experiencing OSA. Methods Geographically informed social indicators were used to define social risk factor clusters via latent class analysis. EHR-wide diagnoses were used as predictors of 5-year incidence of major adverse CV events (MACE) using STREAMLINE, an end-to-end rigorous and interpretable automated machine learning pipeline. Results Analyses among over 1.4 million individuals revealed three major social risk factor clusters: lowest (35.7%), average (43.6%) and highest (22.7%) social burden. In adjusted analyses, those experiencing highest social burden were less likely to have received a diagnosis of OSA when compared to those experiencing lowest social burden (OR [95%CI]=0.85[0.82-0.88]). Among those with OSA and free of prior CV diseases (N=4,405), performance of predicting incident MACE reached a AUC of 0.70 overall but varied when assessed within each social risk factor cluster. Feature importance also revealed that different clinical factors might explain predictions among each cluster. Conclusion Results suggest relevant health disparities in the diagnosis of OSA and across clinical predictors of CV diseases among those with OSA, across social risk factor clusters, indicating that tailored interventions geared toward minimizing these disparities are warranted. Support (if any) This research was, in part, funded by the AIM-AHEAD program, National Institutes of Health (NIH) Agreement NO. 1OT2OD032581-01. ADVANCE’s participation in PCORnet® is funded through the PCORI Award RI-OCHIN-01-MC.

  PublisherOA PDFDOI

• Identifying HLA amino acid mismatches that predict kidney graft failure with cross validation using FIBERS 2.0 machine learning binning

  Human Immunology · 2025-09-01

  article1st authorCorresponding
  PublisherDOI

• Impact of HLA class I amino acid positions that distinguish allele lineages on adjusted hazard ratios of graft failure

  Human Immunology · 2025-09-01

  article
  PublisherDOI

Frequent coauthors

• Jason H. Moore

  Cedars-Sinai Medical Center

  85 shared

• Randal S. Olson

  22 shared

• Moshe Sipper

  13 shared

• John H. Holmes

  University of Pennsylvania

  13 shared

• Will N. Browne

  Queensland University of Technology

  12 shared

• Nadine Al‐Naamani

  University of Pennsylvania

  12 shared

• William La Cava

  Harvard University Press

  10 shared

• Danielle L. Mowery

  University of Pennsylvania

  10 shared

Education

• Ph.D., Genetics

  Dartmouth College

  2012

• Masters of Engineering, Biological Engineering

  Cornell University

  2005

• Bachelors of Science, Biological Engineering

  Cornell University

  2004

Awards & honors

• Best paper award in the EvoBIO track (2016)
• Best paper award in the Evolutionary Machine Learning Track…
• Solving the extremely complex 135-bit benchmark multiplexer…