Past Usage: (a) Rögnvaldsson, You and Garwicz (2015) 'State of the art prediction of HIV-1 protease cleavage sites', Bioinformatics, vol 31 (8), pp. 1204-1210. - Discusses the characteristics of the four data sets. - It is shown that a linear SVM with standard orthogonal encoding is the best predictor when tested across data sets. The AUC (Area Under the ROC Curve) values are listed below: Test data 746 1625 Schilling Impens Train data 746 - 0.982 0.870 0.833 1625 0.981 - 0.855 0.811 Schilling 0.933 0.935 - 0.911 Impens 0.902 0.894 0.929 - The 746 and 1625 data sets share many patterns. (b) Kontijevskis, Wikberg and Komorowski (2007) 'Computational Proteomics Analysis of HIV-1 Protease Interactome'. Proteins: Structure, Function, and Bioinformatics, 68, 305–312. Introduced the 1625 data set and used rough set theory on it. Note: the 1625 data set used by Kontijevskis et al. is not identical to the 1625 data set here, there is one octamer that has a different label (same octamer as for the 746 data set). (c) You, Garwicz and Rögnvaldsson (2005) 'Comprehensive Bioinformatic Analysis of the Specificity of Human Immunodeficiency Virus Type 1 Protease'. Journal of Virology, 79, 12477–12486. Introduced the 746 data set and explored linear classifiers plus representations based on amino acid properties. Note: the 746 data set used by You et al. is not identical to the 746 data set here, there is one octamer that has a different label (same octamer as for the 1625 data set). Other relevant information The data sets have appeared in earlier publications (see references in Rögnvaldsson et al. 2015). However, corrections have been made to them here. Details on references where data has been collected are provided for the 746 and the 1625 data set. The origins of the Schilling data and the Impens data are described in the paper Rögnvaldsson et al., 2015.

No