Adeno-associated virus (AAV) vectors have been proved clinically effective for gene therapy of several diseases. In order to develop optimal delivery according to the specific target, current research is focusing on modifications that improve viral capsids. Directed Evolution (DE) is a protein engineering technique widely used in experimental studies. It mimics the process of natural selection: given an initial library of variants with a certain diversity, DE iteratively enriches vectors that successfully achieve a defined goal and ultimately allows convergence towards one or a few best performing variants. Recently, DE has begun to benefit from the massive data production enabled by Next-generation sequencing (NGS). This thesis project is based on the experimental data obtained by recent studies of Planul and Dalkara (2019) at Sorbonne Université. With the ultimate goal of efficiently treating inherited retinal degenerations, their research focuses on gene therapy mediated by non-invasive vector delivery. They performed 6 consecutive rounds of DE on post-mortem human tissue to effectively transfect the outer retinal layers, starting with random peptide-display library of vectors, synthesized inserting random sequences of 7 amino acids (7mer) between positions 587/588 of the AAV2 capsid. Specifically, the viral variants production from the correspondent plasmid library is profoundly influenced by the stability of the modified capsid: in fact, not all the possible inserted sequences result into viable viruses. This preliminary filtering is fundamental for DE since it limits the selection to functional vectors present in the viral library. What factors determine certain inserted sequences to be beneficial or detrimental for viral formation? How can comprehension of this complex biological process be improved? This thesis proposes different Machine learning approaches applied to massive NGS data to answer such questions, providing a characterisation of the stability of the AAV capsid after a 7mer insertion.
I virus adeno-associati (AAV) si sono dimostrati vettori clinicamente efficaci per la terapia genica di diverse malattie. Al fine di sviluppare un rilascio ottimale in base al target specifico, la ricerca attuale si sta concentrando su modifiche che migliorano i capsidi virali. La Directed Evolution (DE) è una tecnica di protein engineering, ampiamente utilizzata negli studi sperimentali. Essa imita il processo di selezione naturale: data una libreria iniziale di varianti con una certa diversità, la DE arricchisce iterativamente i vettori che raggiungono con successo un obiettivo predefinito e alla fine consente la convergenza verso una o poche varianti migliori. Recentemente, la DE ha cominciato a beneficiare della produzione di dati massiva permessa dal Next-generation sequencing (NGS). Questo progetto di tesi si fonda sui dati sperimentali ottenuti dagli studi di Planul e Dalkara (2019) presso la Sorbonne Université. Con l'obiettivo finale di trattare in modo efficiente le degenerazioni retiniche ereditarie, la loro ricerca si concentra sulla terapia genica mediata dal rilascio non invasivo di vettori virali. Essi hanno svolto 6 cicli consecutivi di DE su tessuto umano post mortem al fine di transfettare efficacemente gli strati retinici esterni, a partire da una libreria iniziale di vettori sintetizzata inserendo sequenze casuali di 7 amminoacidi (7mer) tra le posizioni 587/588 del capside AAV2. Nello specifico, la produzione di varianti virali dalla corrispondente libreria di plasmidi è profondamente influenzata dalla stabilità del capside modificato: infatti non tutte le possibili sequenze inserite danno origine a un virus effettivo. Tale filtraggio preliminare è fondamentale per la DE poiché circoscrive la selezione a vettori funzionali presenti nella libreria virale. A causa di quali fattori determinati peptidi inseriti hanno un impatto positivo o negativo sulla formazione virale? Come si può raggiungere una comprensione approfondita di questo complesso processo biologico? Il lavoro di tesi propone diversi approcci di Machine learning applicati a dati massivi di NGS per rispondere a tali domande, fornendo una caratterizzazione della stabilità del capside dell'AAV dopo l'inserimento di un 7mer.
Characterising the stability of genetically modified viral capsids towards retinal gene therapy
MOLINARI, GIULIA
2019/2020
Abstract
Adeno-associated virus (AAV) vectors have been proved clinically effective for gene therapy of several diseases. In order to develop optimal delivery according to the specific target, current research is focusing on modifications that improve viral capsids. Directed Evolution (DE) is a protein engineering technique widely used in experimental studies. It mimics the process of natural selection: given an initial library of variants with a certain diversity, DE iteratively enriches vectors that successfully achieve a defined goal and ultimately allows convergence towards one or a few best performing variants. Recently, DE has begun to benefit from the massive data production enabled by Next-generation sequencing (NGS). This thesis project is based on the experimental data obtained by recent studies of Planul and Dalkara (2019) at Sorbonne Université. With the ultimate goal of efficiently treating inherited retinal degenerations, their research focuses on gene therapy mediated by non-invasive vector delivery. They performed 6 consecutive rounds of DE on post-mortem human tissue to effectively transfect the outer retinal layers, starting with random peptide-display library of vectors, synthesized inserting random sequences of 7 amino acids (7mer) between positions 587/588 of the AAV2 capsid. Specifically, the viral variants production from the correspondent plasmid library is profoundly influenced by the stability of the modified capsid: in fact, not all the possible inserted sequences result into viable viruses. This preliminary filtering is fundamental for DE since it limits the selection to functional vectors present in the viral library. What factors determine certain inserted sequences to be beneficial or detrimental for viral formation? How can comprehension of this complex biological process be improved? This thesis proposes different Machine learning approaches applied to massive NGS data to answer such questions, providing a characterisation of the stability of the AAV capsid after a 7mer insertion.| File | Dimensione | Formato | |
|---|---|---|---|
|
2020_10_Molinari.pdf
non accessibile
Descrizione: testo tesi
Dimensione
3.87 MB
Formato
Adobe PDF
|
3.87 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/166801