Trained interactions can discriminate concentration patterns across nearly 1,000 molecules

April 9, 2022

(left) Trained binding interactions can assemble multiple distinct structures from a soup of molecules through distinct nucleation pathways. Concentration patterns, in which some molecules are more abundant than others, preferentially select one nucleation pathways over others. Consequently, the molecular soup can be trained to classify high-dimensional concentration patterns by assembling distinct structures in response to classes of concentration patterns. (right) Experimental realization of such a `molecular neural network’ using 917 molecules (DNA oligomers) that assemble distinct structures in response distinct concentration patterns.


Murugan lab, MRSEC funded grad student: Jackson O’Brien, External collaborator: Winfree lab, Caltech