The goal of this paper is a first attempt to link the geology and mineral deposits of the Gondwana GIS, called GO-GEOID, to spatial statistics and map modelling. Such integration allows, for example, meaningful identification of the mineral potential of unexplored regions relative to those around them that are known to host major ore deposits. We use gold and tin deposits of Africa to derive prescriptive and predictive potential mineral maps. In the case of the prescriptive output, techniques such as Boolean algebra or multicriteria analysis are explored together with the expert knowledge of the geologist (knowledge-driven models). In the case of the predictive output/models, map layers are combined using a Bayesian probabilistic framework; these models become data-driven and are more objective. Confidence in our predictive model for gold mineralisation is provided through validation against recently discovered gold deposits in Tanzania.