Center for Functional Nanomaterials Seminar

Molecular simulations are becoming indispensable tools for designing and characterizing soft matter systems, such as proteins, nucleic acids, polymers, and surfactants. In this talk, I will present a combined simulation-experimental approach toward optimized DNA biosensor design. We developed a coarse-grained model for simulating DNA hybridization on surface. Simulations performed using this model reveals presence of conformational heterogeneities corresponding to partially hybridized structures on the surface, which results in false positives and false negatives during sequence detection. Presence of such conformational heterogeneities was later confirmed with wet lab experiments. We propose a solution of this problem by customizing the sensor surface area according to the molecular dimension, which will increase the diagnostic accuracy of hybridization-based DNA sequence detection methods.

In the final part of the talk, I will discuss how advanced machine learning techniques can be used to tackle the emerging "big data" problem in molecular simulation research. With recent advances of computing power, emerging efficient sampling schemes, and wide interest in soft matter simulations, the cumulative amount of simulation data is becoming massive. How to merge, analyze, and build predictive models from this massive amount of data thus becomes a research challenge. I will demonstrate how machine learning techniques enable better understanding of large-scale, heterogeneous protein simulation data.