Malaria claims millions of lives annually. This global killer causes approximately 2.7 million annual deaths worldwide; addressing this problem has become more and more crucial. Due to pathogen evolution no efficient vaccine for treatment of malaria currently exists. As infection has developed as a field of study, it became ever more clear that infections could only be understood within the context of the host-pathogen community. This project aims to predict possible drug targets based on host-pathogen interactions rather than just protein-protein interactions within a single organism. Similar to Lee et al. (2008) pathogen-host interaction predictions are based on orthology, these interactions are then analysed to identify potential drug targets. This could potentially aid researchers in their continuous battle against malaria and the larger scale battle against pathogen evolution. To predict in vitro host-pathogen interactions DISCOVERY uses an ortholog clustering method called ORTHOMCL. ORTHOMCL is very suitable for ortholog clustering of malaria data for two reasons. Firstly, it is capable of distinguishing between recent paralogs and ancient paralogs, which enables the inclusion of recent paralogs together with orthologs. Secondly, ORTHOMCL was initially developed for the use of malaria data. Identification of in vitro interactions is followed by scoring methods to determine the possible in vivo interactions that might occur between the Plasmodium parasite and the human and mosquito hosts. Scoring measures and weights were applied to 5 different factors to calculate a final score. These final scores allow user input to define the preferred stringency when viewing possible interactions with a single protein. These different factors are sequence similarity, PEXEL/VTS motif presence, microarray expression, metabolic map sharing and sub-cellular locations boundaries. DISCOVERY’S results and results from two other (Dyer et al. and Lee et al.) in silico prediction methods were compared with Vignali et al’s experimental interactions which are based on a yeast two-hybrid approach. Similar to results shown by Doolittle and Gomez these comparisons had poor results. The next step was to compare the in silico results with each other. Dyer et al’s and Lee et al‘s results compared poorly with each other. Although DISCOVERY did not compare well with Dyer et al’s results, comparisons with Lee et al. showed more promise. Poor comparisons with Dyer et al. may be due to their unique approach to predict in vitro host-pathogen interactions. This project identified the lack of enough valid and reliable experimental data to evaluate in silico prediction methods as a definite challenge for host-pathogen interaction predictors. Although this is a major problem, DISCOVERY improved on older prediction methods with the use of a more applicable ortholog clustering technique and the use of more assessment methods during in vivo interaction predictions. DISCOVERY also used scoring methods rather than exclusion methods during the identification of in vivo interactions. This allows a user to specify a threshold of sensitivity when viewing interactions. The true potential of host-pathogen interaction predictions would only be realized when the gap between predictions and evaluation data is bridged.