A prediction has been made of the Charpy ductile-brittle transition temperature at high irradiation levels from a dataset of 459 low-activation ferritic/martensitic steels. It follows a similar study of the yield stress of some 1811 similar alloys (Windsor et al , Modelling Simul. Mater. Sci. Eng. 16 (2008) 025005) Neural networks have previously been used by Cottrell et al (Journal of Nuclear Materials 367-370, 2007, 603-609) to model the Charpy ductile-brittle transition temperature of these alloys. The same dataset has been used in this study, but has been divided into a training set containing the majority of the dataset with low irradiation levels, and a test set which contains just those alloys which have been irradiated above a given level. For example, some 7.2% of the dataset was irradiated above 20 dpa. Good predictions were made using two different codes and methods. As with the yield stress prediction, the target-driven components method, where linear combinations of the atomic inputs are chosen to reduce the test residual, achieved the best results.