Research > Research Areas
Cancer Informatics and Precision Medicine
Cancer informatics and precision medicine leverage genomic, molecular, imaging and clinical data to support more precise diagnosis, risk stratification, and treatment selection across the cancer care continuum. By integrating diverse data sources, from tumor sequencing and pathology to longitudinal clinical outcomes, these approaches enable clinicians to tailor therapies to the biological characteristics of each patient’s disease.
Our work spans the cancer care continuum, applying artificial intelligence and computational methods to causal discovery, biomarker identification, early detection, metastatic risk prediction, clinical decision support and visual analytics. Anchored in close collaboration with clinicians at UPMC Hillman Cancer Center and partners across academia and industry, this work translates computational advances into tools and insights that are biologically meaningful and clinically actionable.
Projects
Developing a novel causal discovery framework to unveil individualized cell-cell communication networks
Differences in cell composition and functional states within individual tissue environments lead to heterogeneity. Due to its complexity, understanding how cells in tissue communicate and influence each other's cellular states and how they eventually reach homeostasis is a challenging problem. The fundamental hypothesis underpinning this project is that a cell-cell communication network (CCCN) is a causal network in which changes in a cell's cellular state can causally influence neighboring cells' states through ligand-receptor (LR) signal transduction. Different tissue samples have distinct cell compositions and thereby communicate through an individualized CCCN. This project develops a computational framework that integrates deep learning and causal discovery algorithms to unveil the individualized (or instance-specific) CCCNs underlying heterogeneity in the tissue environment of individual tissue samples. The versatility of this framework enables its application to the study of CCCNs in both normal physiological conditions and various disease microenvironments.
Faculty: Xinghua Lu, MD, PhD, Gregory F. Cooper, MD, PhD, and Lujia Chen, PhD
Collaborator: Lokesh Sharma, PhD
Sample Publication: Chen X, Chen L, Kürten CHL, Jabbari F, Vujanovic L, Ding Y, Lu B, Lu K, Kulkarni A, Tabib T, Lafyatis R, Cooper GF, Ferris R, Lu X. An individualized causal framework for learning intercellular communication networks that define microenvironments of individual tumors. PLoS Comput Biol. 2022 Dec 22;18(12):e1010761.
Biomarker discovery for VEGFR multi-kinase inhibitor and immune checkpoint inhibitor combination therapy in advanced-stage microsatellite stable (MSS) colorectal cancer
According to the American Cancer Society, colorectal cancer (CRC) is the third most diagnosed cancer in the United States, with approximately 152,000 new cases each year. Immune checkpoint inhibitors (ICIs) have demonstrated clinical efficacy in patients with microsatellite instability–high (MSI-H) CRC, and four ICIs have been approved by the U.S. Food and Drug Administration for the treatment of advanced MSI-H disease. However, ICIs show limited efficacy in patients with microsatellite-stable metastatic CRC, with responses typically rare. This project develops an integrative multi-omics analytical framework to identify predictive biomarkers for a promising combination therapy comprising the multi-tyrosine kinase inhibitor cabozantinib and a PD-L1–targeting ICI. Building on recent evidence of clinical activity in a subset of CRC patients, the project employs advanced computational approaches, including deep learning, topic modeling and causal discovery, to derive comprehensive and biologically interpretable representations of cellular states within the tumor microenvironment, which are then leveraged to predict therapeutic response and uncover mechanisms underlying treatment sensitivity.
Faculty: Xinghua Lu, MD, PhD
Collaborator: Anwaar Saeed, MD
Sample Publication: Zhang H, Lu B, Cooper GF, Saeed A, Paisley JW, Lu X, Chen L. Deconvoluting single-cell transcriptomics reveals cellular programs regulated by cell-cell communication in colorectal cancer. bioRxiv [Preprint]. 2025 Apr 15.
Cancer systems biology and early cancer detection
This project develops machine learning frameworks that integrate genomic, molecular and spatial data to identify cancer-specific biomarkers for early detection. The goal is to produce signatures that are biologically meaningful, clinically actionable and sensitive enough to power the next generation of non-invasive screening.
Faculty: Hatice Ulku Osmanbeyoglu, PhD
Collaborators: Carnegie Mellon University, University of Pittsburgh, C-DETECT Consortium, industry partners
Transcriptomics and cancer biology
Research in this area investigates the molecular mechanisms underlying breast cancer progression, with particular emphasis on the transcriptional consequences of growth factor signaling and the roles of long non-coding RNAs (lncRNAs) in cancer biology. A central focus is the regulation of SNHG7, an oncogenic lncRNA controlled by insulin-like growth factor 1 (IGF1) signaling through a MAPK-dependent negative feedback mechanism. Disruption of this regulatory circuit promotes unconstrained proliferation in a subset of breast cancers, positioning IGF1-regulated non-coding RNA networks as candidates for therapeutic targeting. This work is complemented by broader investigations into IGF1 and insulin receptor biology and the landscape of lncRNA dysregulation across breast cancer subtypes. The group also contributes translational genomics and transcriptomic expertise to collaborative studies spanning multiple cancer types, including investigations of profilin-1 and MRTF signaling in breast cancer cell migration and vascular pathology, androgen receptor expression in osteosarcoma lung metastasis, and intragenic rearrangement burden as a determinant of immune cell infiltration and response to immune checkpoint blockade. Together, these projects reflect an integrative approach to cancer biology that bridges computational and experimental methods to advance understanding of tumor progression and treatment response.
Faculty: David N. Boone, PhD
Collaborators: Adrian V. Lee, PhD, Steffi Oesterreich, PhD, Uma Chandran, MPH, PhD, MSIS, Partha Roy, PhD, and David Gau, PhD
Sample Publications: Boone DN, Warburton A, Som S, Lee AV. SNHG7 is a lncRNA oncogene controlled by insulin-like growth factor signaling through a negative feedback loop to tightly regulate proliferation. Scientific Reports. 2020;10:8583.
Chawla P, Sharma I, Gau D, Eder I, Chen F, Yu V, Welling N, Boone D, Taboas J, Lee AV, Larregina A, Galson DL, Roy P. Breast cancer cells promote osteoclast differentiation in an MRTF-dependent paracrine manner. Mol Biol Cell. 2025 Jan 1;36(1):ar8. doi: 10.1091/mbc.E24-06-0285. Epub 2024 Dec 4. PMID: 39630611; PMCID: PMC11742114.
Personalized prediction of metastatic breast cancer
Breast cancer patients may remain at risk for metastatic recurrence years after initial diagnosis and treatment, making individualized predictions important for guiding follow-up care and avoiding under- or overtreatment. This project develops artificial intelligence and machine learning methods to predict the later occurrence of breast cancer metastasis using real-world clinical data. The work integrates Bayesian network methods, risk factor discovery, deep learning and grid search strategies to identify direct and interactive risk factors and predict 5-, 10-, and 15-year metastatic risk. By combining interpretable probabilistic modeling with deep learning, this research aims to improve risk stratification and support precision oncology for breast cancer patients.
Faculty: Xia Jiang, Ph.D.
Collaborators: Alan Wells, MD, DMSc, Adam Brufsky, MD, PhD, Chuhan Xu, Yijun Zhou, and Gomez Marti
Sample Publication: Jiang X, Xu C. Deep learning and machine learning with grid search to predict later occurrence of breast cancer metastasis using clinical data. J Clin Med. 2022;11(19):5772.
Clinical decision support and real-world data analytics for breast cancer
This project develops clinical decision support, natural language processing, and machine learning approaches to support personalized breast cancer assessment and treatment planning using clinical, genomic and real-world data. The work integrates patient-specific information, including clinical features, genomic profiles, diagnostic subtype, tumor stage and grade, comorbidities, and outcomes, to assist physicians in assessing patient conditions and making individualized treatment recommendations. Related work also uses natural language processing and machine learning to identify breast cancer recurrence from clinical data. Together, these efforts aim to improve recurrence detection, prognosis, and treatment selection in precision oncology.
Faculty: Xia Jiang, PhD
Collaborators: Alan Wells, MD, DMSc, Adam Brufsky, MD, PhD, Richard E. Neapolitan, PhD, Zhexian Zeng, PhD, Xia Li, Sijia Espino, Ajay Roy, and Shah Khan
Sample Publication: Jiang X, Wells A, Brufsky A, Neapolitan RE. A clinical decision support system learned from data to personalize treatment recommendations towards preventing breast cancer metastasis. PLoS One. 2019;14(3):e0213292.
Breast Cancer Research Foundation-funded AURORA project: multi-omics profiling of metastatic breast cancer
This project conducts molecular studies of primary and metastatic breast cancer to discover mechanisms of disease initiation, progression, and response to therapy. Biospecimens are profiled using RNA-seq, whole-exome sequencing, whole-genome sequencing and methylation platforms, and data sets include molecular, clinical and imaging data. Key consortium participants include the University of North Carolina, Mayo Clinic, Nationwide Children's Hospital, Washington University, Van Andel Institute, and the University of Pittsburgh. The Department of Biomedical Informatics' Cancer Bioinformatics Services (CBS), along with the Pittsburgh Supercomputing Center, leads the data coordination center, which ingests data; collects metadata; and annotates, harmonizes and shares data. CBS also assists with data analysis at the request of the molecular working group. AURORA consists of two phases: retrospective and prospective. Analysis of the retrospective cohort is complete and has been published, and data have been shared via multiple portals; the prospective analysis and the combined retrospective-prospective analysis are ongoing.
Faculty: Uma Chandran, MPH, PhD, MSIS
Collaborators: Adrian Lee, PhD, Charles Perou, PhD, Elaine Mardis, PhD, Katherine Hoadley, PhD, and Peter Laird, PhD
Sample Publication: J Garcia-Recio S, Hinoue T, Wheeler GL, et al. Multiomics in primary and metastatic breast tumors from the AURORA US network finds microenvironment and epigenetic drivers of metastasis. Nat Cancer. 2023;4(1):128–147.
Immunogenomics of uveal melanoma to predict therapy response
Uveal melanoma has traditionally been considered an "immune-cold" tumor. This project uses transcriptomics to characterize its transcriptional landscape and identify predictors of therapy response. The Cancer Bioinformatics Services conducted a bulk RNA-seq transcriptional analysis, integrating it with clinical and immune data to identify biomarkers and signatures associated with treatment response.
Faculty: Uma Chandran, MPH, PhD, MSIS
Collaborator: Udai Kammula, PhD
Sample Publication: Leonard-Murali S, Bhaskarla C, Yadav GS, et al. Uveal melanoma immunogenomics predict immunotherapy resistance and susceptibility. Nat Commun. 2024;15(1):2863.
Cancer visual analytics
This work develops human-centered data visualization and interactive analytic tools that make complex cancer data more accessible and actionable. Drawing on expertise in human-computer interaction and biomedical informatics, the focus is on building highly interactive, user-centered systems for finding and exploring biomedical datasets spanning cancer genomic data, clinical records and patient histories. A central project is DeepPhe, an National Cancer Institute-funded initiative that extracts longitudinal patient histories from clinical notes using natural language processing and pairs them with visual analytics tools to support cohort discovery. By grounding tool design in qualitative inquiry into clinicians' information needs and goals, this work bridges the gap between raw data and meaningful insight, enabling researchers and clinicians to explore cancer genomic and clinical data to support discovery and precision oncology decision-making.
Faculty: Harry Hochheiser, PhD
Collaborators: Jeremy L. Warner, MD, MS, and Guergana K. Savova, PhD
Sample Publication: Yuan Z, Finan S, Warner J, Savova G, Hochheiser H. Interactive exploration of longitudinal cancer patient histories extracted from clinical text. JCO Clin Cancer Inform. 2020 May;4:412-420.