Output details
11 - Computer Science and Informatics
Newcastle University
Contact map prediction using a large-scale ensemble of rule sets and the fusion of multiple predicted structural features
<28> This paper builds on results from EPSRC project GR/T07534/01. Our a new machine intelligence algorithm has 3 key innovations: new ensemble architecture for data sparseness and large training sets; new information fusion techniques and our own new metric called Recursive Convex Hull; robust rule learning system designed large dimensionality spaces. Independently evaluated by CASP8 (2008) and CASP9 (2010) organisers, the bi-annual international competition that benchmarks state-of-the-art methods for predicting protein structures - a biological holy grail- our method was ranked in the top 3 and best for contact map prediction in brand new folds.