Predictive Models for Gene Expression Engineering
Restricted (Penn State Only)
- Author:
- Lafleur, Travis
- Graduate Program:
- Chemical Engineering
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- July 31, 2023
- Committee Members:
- Andrew Zydney, Major Field Member
Howard Salis, Chair & Dissertation Advisor
Edward O'Brien, Outside Unit & Field Member
Costas Maranas, Major Field Member
Robert Rioux, Professor in Charge/Director of Graduate Studies - Keywords:
- Transcription
Machine Learning
Biological Engineering
Synthetic Biology - Abstract:
- Massively parallel reporter assays (MPRAs) leverage state of the art DNA synthesis and sequencing technologies to test thousands of defined biological experiments in a single test tube. In this dissertation, I discuss the progress we have made towards automating MPRA design and analysis with particular emphasis on the use of MPRAs for transcription rate measurements. I demonstrate the utility of these automated approaches through the design, characterization, and analysis of thousands of promoter sequence variants. Then, using that data, I show how predictive transcription rate models can be trained using a hybrid modeling approach that combines statistical thermodynamics and machine learning formalisms. Importantly, I show how this approach has enabled the training of a transparent model with interpretable coefficients which has enhanced our understanding of transcription dynamics and regulatory mechanisms. Finally, I will expand these experimental and computational approaches to five additional σ factors to create an algorithm suite capable of designing and debugging genetic systems. Throughout the dissertation, I will showcase the algorithms’ applications in automating forward engineering of promoters with target expression rates, automating engineering of orthogonal promoters, and the accurate identification of unintended transcriptional start site locations within large synthetic genetic systems.