Work

Bringing Models to the Domain: Deploying GaussianProcesses in the Biological Sciences

Year: 2017
Type: dissertation
Abstract: Recent developments in single cell sequencing allow us to elucidate \nprocesses of individual cells in unprecedented detail. This detail \nprovides new insights into the progress of cells during cell type \ndifferentiation. Cell type heterogeneity shows the complexity of cells \nworking together to produce organ function on a macro level. The \nunderstanding of single cell transcriptomics promises to lead to the \nultimate goal of understanding the function of individual cells and \ntheir contribution to higher level function in their environment. \n \nCharacterizing the transcriptome of single cells requires us to \nunderstand and be able to model the latent processes of cell functions \nthat explain biological variance and richness of gene expression \nmeasurements. In this thesis, we describe ways of jointly modelling \nbiological function and unwanted technical and biological confounding \nvariation using Gaussian process latent variable models. In addition \nto mathematical modelling of latent processes, we provide insights \ninto the understanding of research code and the significance of \ncomputer science in development of techniques for single cell \nexperiments. \n \nWe will describe the process of understanding complex \nmachine learning algorithms and translating them into usable \nsoftware. We then proceed to applying these algorithms. We show how \nproper research software design underlying the implementation can lead \nto a large user base in other areas of expertise, such as single cell gene \nexpression. To show the worth of properly designed software underlying \na research project, we show other software packages built upon the \nsoftware developed during this thesis and how they can be applied to \nsingle cell gene expression experiments. \n \nUnderstanding the underlying function of cells seems within reach \nthrough these new techniques that allow us to unravel the \ntranscriptome of single cells. We describe probabilistic techniques of \nidentifying the latent functions of cells, while focusing on the \nsoftware and ease-of-use aspects of supplying proper research code to \nbe applied by other researchers. \n \n
Language: English
Open Access status: green
OpenAlex ID: W2769077185
Cites: 78
Cited by: 1
Related to: 9