Genotypic Predictors of Human Immunodeficiency Virus Type 1 Drug Resistance
Rhee SY, Taylor J, Wadhera G, BenHur A, Brutlag DL, Shafer RW.,
Proceedings of National Academy of Sciences of the United States of America, Oct 25, 2006.
Data
 Sets of mutations [Table 1]:
 Genotypephenotype correlation data sets:
 Phenotype cutoffs between susceptible, low/intermediate resistance and highlevel resistance for each drug
Results
 Summary of 10 x 5 foldcross validation results:
Tables
 Prediction accuracy [Table 3]
 Number of highly discordant predictions according to mutation set and regression method
[Table 4]
 Highly Discordant Isolates: Predictions based on Least Angle Regression (LARS) Using Nonpolymorphic Treatment Selected Mutations
[Table 5]
 Correlation coefficients (r2) between actual and predicted result according to regression method and mutation dataset
[Table 6]
 Mean squared error between actual and predicted result according to regression method and mutation dataset
[Table 7]
 Regression coefficients for the PI treatmentselected mutations [Table 8],
the NRTI treatmentselected mutations [Table 9] and the NNRTI treatmentselected mutations [Table 10] in the Least Squares Regression (LSR),Support Vector Regression (SVR), and Least Angle Regression (LARS) Models
Figures
 Leastsquares regression coefficients for protease inhibitor [figure 1A], nucleoside RT inhibitor [figure 1B] and nonnucleoside RT inhibitor [figure 1C] treatmentselected mutations
 Distribution in the number of nonpolymorphic TSMs [figure 2A] and expert panel mutations [figure 2B] per isolate
 Rate of learning using regression and nonpolymorphic treatmentselected mutations (TSMs) [figure 3]
 Regression coefficients for each of the PI, NRTI and NNRTI TSMs determined by different regression methods [figure 4]
[figure legends]
