In this talk some of the research being carried out in the UCL Bioinformatics Group will be discussed, with a focus on our work on protein structure prediction. The so-called "protein folding problem" is perhaps the biggest of the remaining challenges in molecular biology and significat progress in this area could have impact in diverse areas such as nanotechnology and even clinical medicine. The puzzle as to how the complex 3-D structure of a protein is encoded purely by its amino acid sequence is being tackled by both experimentalists and theoreticians alike. Although a full understanding of how proteins self-assemble into their native structures is a long way off, a more tractable sub-problem is the challenge of programming a computer to predict the structure of a given protein from its sequence. In computational terms, the protein folding problem can be described as a "black box" into which a string of 20 different letters is fed and a set of X,Y,Z coordinates for each atom in the protein produced as the output. In this talk, I will review some of the work from my own lab in developing computationally efficient methods for simulating protein folding, including some recent experiments in the use of modern graphics hardware to allow bench scientists to fold proteins on their desktop PCs, and the use of machine learning methods to produce ways of discriminating correctly folded from misfolded proteins.