Nanoparticles (NPs) with controlled surface modification have unique properties and broad applications, ranging from biomedicine to catalysis. However, surface modification processes usually generate a mixture of NPs with a range of different surface structures. Herein, based upon coarse-grained molecular simulations, we propose a facile two-step approach for recognition of targeted nanoparticles (TNPs) by a second type of nanostructure, followed by selective precipitation driven by an external force, to extract TNPs from NP mixture with similar surface structures. As an example, we consider mixtures of partially functionalized nanosheets representing graphene flakes of uniform size and shape but varying surface modification. Our results show that Janus particles can effectively recognize and selectively adsorb on targeted graphene NPs (TGNPs) to allow their separation from NP mixture. In our molecular simulations, high purity TGNPs (up 99.9%) could be extracted from a mixture containing only similar to 10% TGNPs after three to four rounds of separation.