A review of mathematical models for the pathogenesis of HIV is presented. Model identifiability and identification techniques are applied using the biomedical model and clinical data. Identifiability analysis aids in the determination of measured values and measurement frequency. Parameter identification methods are chosen and developed for sparse and rough samples. Results are reported on two case studies: vaccine readiness in Southern Africa, drug effectiveness and therapy failures on existing patients in France. Ongoing research programmes and future opportunities are considered. (c) 2006 Elsevier Ltd. All rights reserved.