The performance of RV-Typer was validated using receiver operating characteristic (ROC) analysis.
The sequences in true positive (TP) and negative (TN) data sets were given as an input to RV-Typer. Based on the serotype predictions for sequences in TP and TN data sets, the sensitivity and specificity values were calculated. They are as following,
The serotypes of all the sequences in TP data set were correctly predicted by RV-Typer. No serotypes were predicted for sequences in TN data set.
In order to assess the efficacy of RV-Typer in typing of recombinant VP1 gene sequences, the intra- and inter-type recombinant sequences of VP1 gene were simulated. The sequences were simulated at varying levels of proportions of major and minor parents such as 90-10, 80-20, 70-30, 60-40 and 50-50% of sequence regions from respective parent types. While compiling the simulated intra- and inter-typic recombinant data sets (100 sequences in each), we retained the equal proportion of sequences (~33%) from RV-A, B and C species.
These simulated sequences were given as an input to RV-Typer server. The results of serotype prediction for simulated recombinant sequences are available here. It can be observed from these tables that RV-Typer assigned the type of major parent to both intra- and inter-typic recombinant sequences having 90-10% and 80-20% of sequence contribution from respective parents. Whereas in case of recombinant sequences having 70-30%, 60-40% and 50-50% of sequence proportion from respective parents, RV-Typer did not assign any serotype to most of the sequences and in few cases it assigned serotype of minor parent. Only ~6% of recombinant sequences were assigned with the type other than parent types. Therefore, in case of inconsistent results between RV-Typer and BLAST report, we suggest users to carry out recombination detection analysis.
The simulated recombinant sequence data can be downloaded from here.