De Hoffer, Adele (2025) ENGINEERING GENE EXPRESSION IN MAMMALIAN CELLS WITH DEEP LEARNING MODELS. [Tesi di dottorato]

	Documento PDF DEHOFFER_ADELE_37_PARZIALE.pdf Download (4MB)
	Documento PDF DEHOFFER_ADELE_37_COMPLETO.pdf Visibile a [TBR] Amministratori dell'archivio Download (11MB) | Richiedi una copia

Abstract

Gene expression regulation in mammalian cells is a complex process that can be modulated at multiple levels. This complexity makes it challenging to achieve precise spatiotemporal control, requiring both dynamic regulatory systems and cell-type–specific regulatory elements. This thesis explores two deep learning–based approaches to controlling gene expression in synthetic biology contexts. The first project develops a data-driven model predictive controller (MPC) that employs a neural network to model and dynamically regulate the expression of an exoge- nous gene controlled by a synthetic, inducible Tet-Off promoter. This approach enables precise temporal modulation of gene expression in response to external stimuli. The second project, PromoterGPT, leverages a generative neural net- work to design synthetic promoters with cell-type specificity across three human cell lines (HepG2, K562, and SK-N-SH), addressing the challenge of engineering DNA sequences that drive selective gene expression. The designed promoters show biological relevance through the enrichment of transcription factor binding motifs consistent with the target cell types. While the two projects focus on distinct aspects of gene regulation—dynamic temporal control versus sequence- driven cellular specificity—both demonstrate the utility of neural networks as versatile tools for modeling and controlling gene expression. Together, these studies advance synthetic biology by providing data-driven strategies for both temporal modulation and cell-type–specific design of gene regulatory elements

Downloads

Actions (login required)

Modifica documento