The 70 kDa heat shock proteins expressed by Plasmodium falciparum (PfHsp70s) are believed to be essential to both the survival and virulence of the malaria parasite. A total of six Hsp70 genes have been identified in the genome of P. falciparum. However, only four of these encode canonical Hsp70s, which are believed to localise predominantly in the cytosol (PfHsp70-1 and PfHsp70-x), the endoplasmic reticulum (PfHsp70-2) and mitochondria (PfHsp70-3) of the parasite. These proteins bind and release peptide substrates in an ATP-dependent manner, with the aid of a J-domain protein cochaperone and a nucleotide exchange factor (NEF). The aim of this study was to identify the residues involved in the interaction of these PfHsp70s with their peptide substrates, their J-domain cochaperones and potential NEFs. These residues were then mapped to three-dimensional (3D) structures of the proteins, modelled in three different conformations; each representing a different stage in the ATPase cycle. Additionally, these proteins were compared to different types of Hsp70s from a variety of different organisms and sequence features found to be specific to each PfHsp70 were mapped to their 3D structures. Finally, a novel modelling method was suggested, in which the structures of templates were remodelled to improve their quality before they were used in the homology modelling process. Based on the analysis of residues involved in interactions with other proteins, it was revealed that each PfHsp70 displayed features that were specific to its cellular localisation and each type of Hsp70 was predicted to interact with a different set of NEFs. The study of conserved features in each PfHsp70 revealed that PfHsp70-x displayed various sequence features atypical of both Plasmodium cytosolic Hsp70s and cytosolic Hsp70s in general. Additionally, residues conserved specifically in Hsp70s of Apicomplexa, Plasmodium and P. falciparum were identified and mapped to the each PfHsp70 model. Although these residues were too numerous to reveal any information of specific value, these models may be useful for the purposes of aiding the design of drug compounds against each PfHsp70. Finally, the novel modelling approach did show some promise. Half of the models produced using the modified templates were of a higher quality than their counterparts modelled using the original templates. This approach does still require a lot of validation work and statistical evaluation. It is hoped that it could prove to be a useful approach to homology modelling when the only templates available are poor quality structures.