Scholes, Harry; (2020) On Computable Protein Functions. Doctoral thesis (Ph.D), UCL (University College London).

Preview

Text harry_scholes_phd_thesis_edited.pdf - Accepted Version Download (11MB) | Preview

Abstract

Proteins are biological machines that perform the majority of functions necessary for life. Nature has evolved many different proteins, each of which perform a subset of an organism’s functional repertoire. One aim of biology is to solve the sparse high dimensional problem of annotating all proteins with their true functions. Experimental characterisation remains the gold standard for assigning function, but is a major bottleneck due to resource scarcity. In this thesis, we develop a variety of computational methods to predict protein function, reduce the functional search space for proteins, and guide the design of experimental studies. Our methods take two distinct approaches: protein-centric methods that predict the functions of a given protein, and function-centric methods that predict which proteins perform a given function. We applied our methods to help solve a number of open problems in biology. First, we identified new proteins involved in the progression of Alzheimer’s disease using proteomics data of brains from a fly model of the disease. Second, we predicted novel plastic hydrolase enzymes in a large data set of 1.1 billion protein sequences from metagenomes. Finally, we optimised a neural network method that extracts a small number of informative features from protein networks, which we used to predict functions of fission yeast proteins.

Download activity - last month

Export as CSVJSONXML

Download activity - last 12 months

Export as CSVJSONXML

Downloads by country - last 12 months

Export as XMLJSONCSV

Archive Staff Only

View Item