Research
Discover further insights into our research
The Silverbush Lab is dedicated to unraveling the complexities of tumor heterogeneity and plasticity. Our goal is to understand the mechanisms through which cancer cells transform, becoming more aggressive or evading treatment.
We work with human samples, often from clinical trials, to uncover how these mechanisms influence the outcome of a clinical trial and how we can leverage these mechanisms to improve outcomes. We place special emphasis on hard-to-treat cancers characterized by high levels of heterogeneity, with a particular focus on brain cancers, known for their notoriety in heterogeneity and aggressiveness.
To achieve our objectives, we develop and utilize a combination of wet lab and computational multi-omic single-cell tools. Our interdisciplinary team is composed of MDs, biologists, immunologists, computational biologists, and experts in computer science, with a strong focus on machine learning (ML) and artificial intelligence (AI). Together, we are committed to advancing our understanding of cancer’s complexity and devising innovative strategies to combat it.
Deciphering Tumor Evolution Using Multi-omic Single Cells
Tumors are complex systems, heterogeneous on multiple levels. We can observe their patterns of heterogeneity at the transcriptomic, genetic, and epigenetic levels. However, we lack an understanding of how these levels work in conjunction or how they contribute to treatment resistance. In this body of work, we develop both wet lab and computational solutions to detect genetic alterations (up to point mutation resolution!), DNA methylation patterns, and gene expression in the same single cells, technologies that exist in only a handful of labs in the world. We then use this data to uncover the various layers of intratumoral heterogeneity and establish connections between them. The goal is to identify genetic and epigenetic factors that drive cancer cell behavior, making them prime targets for treatments. These treatments include attacking aggressive cancer cells directly and differentiation therapies, which involve coaxing cancer cells into a weaker state that can be more effectively targeted with combined therapies.
Decoding Clinical Trial Outcomes in Real Time
For decades, teams have attempted to discover effective treatments for GBM and have tested therapeutics within the framework of clinical trials. To date, no trial has shown significant improvement, but recent efforts at UPenn and three other centers are showing great promise for glioma therapy. In this body of work we process and analyze the valuable samples from dedicated patients undergoing immunotherapy at UPenn Hospital, examining longitudinal samples throughout the trial. We employ single-cell technologies and create immune-specific single-cell methods to identify changes in tumor and immune cells. This approach allows us to determine when the therapy is effective, when it falls short, and how we can intervene to prevent it from failing.
ML, Linear Algebra, and AI to Enhance Diagnosis and Prognosis
Every month, we receive an email from our esteemed clinical collaborators requesting assistance with patient diagnoses—a task we are very proud to fulfill. An accurate diagnosis significantly influences treatment plans and is, of course, of paramount importance to the patient. But why can diagnosing a patient be so complicated? The challenge often lies in the bulk samples used for diagnosis, which lump together various heterogeneous malignant cell states and large fractions of microenvironment cells, blending all signals. We develop ML, Linear Algebra, and AI solutions to separate the microenvironment from malignant components in bulk samples. Our approach allows us to closely examine each component, identifying the composition of malignant and immune cells with single-cell precision for the first time. We use these tools not only to aid in diagnosis but also to analyze past cohort and clinical trial data involving bulk samples. Collaborating with Hamburg Medical Hospital, we’ve successfully linked survival and health benefits to tumor composition. Currently, we’re partnering with other teams to connect specific therapies with tumor compositions. Additionally, we’re collaborating with the NIH to provide clinicians direct access to these tools, enabling them to incorporate tumor composition data into their diagnostic and prognostic decisions.