LONDON, UK; 22nd February 2012
The RDI have today installed some new computational models in the HIV-TRePS system and have also introduced an automated sequence upload and mutation extraction function.
The new random forest models, which will power the predictions for cases entered with a genotype, were developed using an expanded training set of 7,263 complete treatment change episodes from more than 30 countries. These are the first models developed that can predict responses to raltegravir. During cross validation the models achieved a mean area under the ROC curve (AUC) of 0.82. When tested with an independent test set of 375 TCEs the combined outputs of the ten models achieved an AUC of 0.87. These figures compare with the AUC values of 0.82 and 0.83 of the previous models during cross validation and independent testing.
The sequence upload function means you can simply add a sequence file or copy and paste the sequence, as a single string or separately for RT and protease, without the need to enter individual mutations. The program automatically checks for sequence length and gaps and extracts the mutations required by the models to make your predictions.
For further information contact:
Andrew Revell (Executive Director, RDI) on +44 207 226 7314, +44 7967 126498 (mobile) or andrewrevell@hivrdi.org
Date published: 22nd February 2012