Non-coding variants, genome regions that are not translated into proteins, are a major cause of genetic diseases, such as Mendelian disorders. The functional effects of these mutations remain difficult to fully comprehend. However, thanks to advances in sequencing technologies — which have greatly enriched biological data banks — and the development of sufficiently powerful hardware, it has become possible to design neural network-based tools capable of analyzing genomic sequences and providing valuable insights into the functional effects of these specific DNA regions. This thesis aims to introduce molecular biology concepts and provide mathematical tools for understanding neural networks. Specifically, it will explore the structure and functioning of convolutional neural networks with the goal of analyzing three tools based on this technology. The thesis will focus on DeepSEA, Basset, and DeepSATA — three tools designed to enhance the understanding of the functional impact of non-coding variants.
Le mutazioni non codificanti, aree del genoma che non vengono codificate in proteine, sono la principale causa dei disturbi genetici, tra cui le malattie mendeliane. Gli effetti funzionali di queste variazioni sono ancora difficili da comprendere completamente. Grazie a nuove tecnologie di sequenziamento — che hanno permesso un significativo arricchimento di banche di dati biologici — e allo sviluppo di supporto hardware sufficientemente potente è stato possibile progettare strumenti basati sulle reti neurali che analizzano sequenze genomiche e forniscono informazioni rilevanti legate agli effetti funzionali di queste particolari mutazioni. Questo elaborato ha l'obiettivo di dare un'introduzione alla biologia molecolare e fornire gli strumenti matematici per comprendere le reti neurali. Più precisamente verrà esplorato il funzionamento e la struttura di una rete neurale convoluzionale con l'obiettivo di riuscire ad analizzare tre strumenti, basati su questa tecnologia. Saranno esaminati DeepSEA, Basset e DeepSATA, tre tool che hanno lo scopo di approfondire gli effetti causati dalle mutazioni non codificanti del genoma.
Reti neurali convoluzionali per lo studio di varianti non codificanti in sequenze genomiche
TRIGOLO, ALESSANDRO
2023/2024
Abstract
Non-coding variants, genome regions that are not translated into proteins, are a major cause of genetic diseases, such as Mendelian disorders. The functional effects of these mutations remain difficult to fully comprehend. However, thanks to advances in sequencing technologies — which have greatly enriched biological data banks — and the development of sufficiently powerful hardware, it has become possible to design neural network-based tools capable of analyzing genomic sequences and providing valuable insights into the functional effects of these specific DNA regions. This thesis aims to introduce molecular biology concepts and provide mathematical tools for understanding neural networks. Specifically, it will explore the structure and functioning of convolutional neural networks with the goal of analyzing three tools based on this technology. The thesis will focus on DeepSEA, Basset, and DeepSATA — three tools designed to enhance the understanding of the functional impact of non-coding variants.File | Dimensione | Formato | |
---|---|---|---|
Trigolo_Alessandro.pdf
accesso aperto
Dimensione
3.65 MB
Formato
Adobe PDF
|
3.65 MB | Adobe PDF | Visualizza/Apri |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/72195