Pattern recognition through molecular self-assembly

The functional role of many weak promiscuous interactions among molecules in biology is not clear and is often assumed to be deleterious. Here, we exploit promiscuous interactions to engineer an experimental system of 917 single-stranded DNA molecules capable of associative pattern recognition on the high dimensional concentration patterns of these molecular assembly components. Such pattern recognition is achieved by exploiting a process of competitive nucleation between different polymorphic DNA structures that are predominantly made of the same molecules but co-localized in different combinations. We test the system with numerous concentration patterns and confirmed nucleation-based pattern recognition through Atomic Force Microscopy (AFM) and fluorescence measurements. We discuss how this system, in conjunction with additional enzymatic components (e.g. DNA ligase), can potentially learn the promiscuous interactions needed to perform unsupervised clustering of concentration patterns.