Project Description:

Gene regulatory networks (GRNs) are the sets of all genes, transcription factors (TFs), and their interactions that determine genetic expression and cell function. We can reconstruct or predict these regulatory networks by utilizing algorithms that take in genetic expression data. Previously, we developed a new tool for GRN reconstruction using single-cell transcriptomic data based on the context likelihood of relatedness (CLR) algorithm, but with added steps to eliminate false edges in the network. Our tool, called Epoch, was benchmarked against CLR as well as GENIE3, another tool commonly used for GRN reconstruction, to compare precision and sensitivity of reconstruction. We demonstrated that, when reconstructing networks from synthetic datasets, all versions of Epoch showed a 5-8 fold area under precision recall curve (AUPR) improvement over random networks, while CLR and GENIE3 showed an approximately 3-fold improvement in AUPR over randomly reconstructed networks. Here, our original Epoch algorithm was translated to Python from R to improve speed and computational power, and to, most importantly, allow for compatibility with the single-cell analysis package SCANPY. The Python version of Epoch was then benchmarked against other competing GRN reconstruction tools as well as beta-tested by graduate students to evaluate its user-friendly nature and convenience. As a more accurate and more efficient GRN reconstruction tool, Epoch works to uncover networks driving developmental pathways and predicts TFs to be dysregulated for the purposes of engineering cell fate.

Project Photo:

Student Team Members

• Jessica Kasamoto

Course Faculty

Project Mentors, Sponsors, and Partners

• Patrick Cahan, MS, PhD
• Emily Su
• Faculty Mentor: Jessica Dunleavey, PhD
• TA Mentor: Sarah Lee, MS