Integration of gene expression data with causal networks

Rajh, Mihael (author), Curk, Tomaž (mentor), Robyn Bleker, Carissa (co-mentor)

With the increased availability of large gene expression datasets comes an increased need for informed methods of data analysis. One class of recent methods involves the use of causal biological networks, which depict causal relationships between molecular events inside the cell. These networks offer the advantage of representing prior biological knowledge in a form that is suited for both computation and human interpretation. However, many of the current methods are held back by implementational challenges, which make them difficult to apply to novel networks. In this thesis, we develop and extend an implementation of the TopoNPA algorithm in the form of a Python package. We present PerturbationX, which features support for custom network syntax, Cytoscape integration, as well as improvements in both edge pruning and permutations. Alongside the implementation, we also provide an estimate of the algorithm's scalability and analyse its sensitivity to noise, missing data, and edge modifications. With the introduction of this robust, open-source tool, we hope to facilitate advancement in the development of causal network algorithms. We aim for the tool to promote insight into experimental data from multiple biological domains.

• Language: English
• Keywords: bioinformatics, causal reasoning, gene expression
• Work type: Master's thesis/paper
• Typology: 2.09 - Master's Thesis
• Organization: FRI - Faculty of Computer and Information Science
• Year: 2023
• PID: 20.500.12556/RUL-152352
• COBISS.SI-ID: 174454531
• Publication date in RUL: 22.11.2023

Secondary language

• Language: Slovenian
• Title: Integracija podatkov o genskem izražanju s kavzalnimi omrežji

Abstract

S povečanjem razpoložljivosti velikih naborov podatkov o izražanju genov se povečuje potreba po informiranih metodah analize podatkov. Eden izmed novejših razredov metod temelji na uporabi kavzalnih bioloških omrežij, ki prikazujejo vzročne odnose med molekularnimi dogodki v celici. Prednost teh omrežij je predstavitev obstoječega biološkega znanja v obliki, ki je primerna tako za računanje kot človeško interpretacijo. Po drugi strani pa številne sorodne metode zavirajo implementacijske težave, ki otežujejo njihovo uporabo z novimi omrežji. V sklopu magistrskega dela smo razvili in razširili implementacijo algoritma TopoNPA v obliki Python programskega paketa. Predstavljamo PerturbationX, ki podpira poljubno sintakso omrežij, integracijo s Cytoscape ogrodjem, ter izboljšave pri odstranjevanju robov in permutacijah. Poleg implementacije podamo tudi oceno skalabilnosti algoritma in analiziramo njegovo občutljivost na šum, manjkajoče podatke ter spremembe robov. Z uvedbo robustnega, odprtokodnega orodja želimo spodbuditi napredek v razvoju algoritmov za kavzalna omrežja. Upamo, da orodje prispeva k spoznanjem o eksperimentalnih podatkih iz različnih področij biologije.

• Keywords: bioinformatika, vzročno sklepanje, izražanje genov