svm-agp is a tool for the detection of atypical genes. On the basis of a one-class SVM it allows for systematic assessment of potentially horizontally transfered genes by evluation of statistical features. In the download section, we provide the python implementation as well as the prediction results on all virus families using different feature sets.