Research
Transfer Learning for Network Biology
Mapping gene networks requires large amounts of data to learn the connections between genes, which impedes discoveries in settings with limited data, including rare diseases and diseases affecting clinically inaccessible tissues. Our lab develops machine learning models that leverage the unprecedented volume of transcriptomic and epigenomic data now available to gain a fundamental understanding of network dynamics that can be democratized to a vast array of downstream applications. We previously developed a novel deep learning model, Geneformer, pretrained on a large-scale corpus of ~30 million single cell transcriptomes to enable context-specific predictions in settings with limited data in network biology through transfer learning. We are now applying Geneformer to accelerate discovery of key network regulators and candidate therapeutic targets for cardiovascular disease. We are also developing approaches of leveraging Geneformer's fundamental understanding of network dynamics to model cell state trajectories and understand the context-dependency of transcription factors in development and disease.
Network-Based Therapeutic Discovery
Mapping the gene regulatory networks driving human disease enables the design of network-correcting treatments that target the core disease mechanism rather than merely managing symptoms. We employ our gene network-based screening approach leveraging machine learning and iPSC disease modeling to identify small molecules with a broad restorative effect on the gene network disrupted in the disease. The machine learning algorithm first learns to distinguish the normal vs. disease state based on the transcriptional signature of the mapped network and then identifies molecules that sufficiently correct the network in the disease cells such that the algorithm classifies them as normal.
Transcription Factor Context-Dependency
Although gene regulatory networks are often represented as static systems, they are dynamic across time and space and highly dependent on cell type, developmental, genetic, and epigenetic contexts. However, the mechanisms of network context-dependency remain poorly understood. Our lab systematically investigates the intrinsic and extrinsic transcriptional and epigenetic modulators of context-dependency of key network regulators using a combination of computational and experimental approaches. We examine how local factors influence the differential sensitivity of specific downstream targets to reduced dosage of transcription factors. We also study how genetic modifiers affect the sensitivity to reduced dosage of transcription factors in varied cell type and developmental contexts to improve our understanding of incomplete penetrance and variable expressivity of genetic disease. Moreover, we determine the temporal hierarchy of gene networks driving disease progression to reveal the most promising therapeutic targets in each stage-specific context.
Using this method, we previously identified a promising network-correcting therapy for cardiac valve disease currently under further investigation towards clinical trial. We are now applying this methodology to additional progressive cardiovascular diseases with no current medical therapies to identify network-correcting therapeutics that can halt disease progression.
Our lab leverages machine learning and experimental genomics to understand the circuitry of gene regulatory networks governing cardiovascular development and how disruption of these networks leads to disease. Investigating the consequences of network rewiring that occurs in disease states uncovers the key mechanisms that coordinate gene transcription to ensure normal development and tissue maintenance. Furthermore, mapping the network dysregulation driving disease allows targeting normalization of central elements to treat the core disease mechanism rather than merely managing symptoms. We apply an innovative network-based framework for therapeutic discovery to cardiovascular disease to accelerate development of much-needed treatments for patients as well as to advance our fundamental understanding of the regulatory circuitry governing human development and disease.
