In this paper, an efficient ghost-cell based immersed boundary method is applied to perform direct numerical simulation (DNS) of mass transfer problems in particle dusters. To be specific, a nine-sphere cuboid duster and a random, generated spherical duster consisting of 100 spheres are studied. In both cases, the duster is composed of active catalysts and inert particles, and the mutual influence of particles on their mass transfer performance is studied. To simulate active catalysts the Dirichlet boundary condition is imposed at the external surface of spheres, while the zero-flux Neumann boundary condition is applied for inert particles. Through our studies, dustering is found to have negative influence on the mass transfer performance, which can be then improved by dilution with inert particles and higher Reynolds numbers. The distribution of active/inert particles may lead to large variations of the duster mass transfer performance, and individual particle deep inside the cluster may possess a high Sherwood number.