On-going research

Molecular self-assembly

Molecular self-assembly is widely being explored as a means to build new functional materials. In previous work we put a lot of effort into simulating the self-assembly of molecules adsorbed on inorganic surfaces, particularly for nanotechnological applications. These simulations were achieved through the use of machine-learned interaction potentials and novel Markov chain Monte Carlo techniques. Representative works include this, this, and this. Our interests are shifting towards bio-active small molecule assemblies, and we particularly want to understand the connection between molecular assembly structure and cellular response. 

(Artwork by Mindy Takamiya 2018)

Bio-active single molecules

Synthetic molecules are often used to control cell functions, however there are few rational strategies for designing them from scratch. We recently developed a new regression-based methodology for designing molecules for inducing cardiac differentiation of iPS cells. It is based on a powerful new molecular feature representation which integrates both molecular conformation and hydrophilicity information, and can be trained using small-sized data sets obtained from chemical screening. See here. We will be expanding this line of work in the future.

Coordination polymers and metal-organic frameworks (MOF)

Coordination polymers and MOFs have fascinating properties which can be fine-tuned through careful choice of metal ions and bridging ligands. In previous work we studied structure-property relationships in coordination polymer semiconductors using first-principles calculations and simplified Hamiltonian models (see here and here). In recent work we used a combination of first-principles modeling and data science to explore how a MOF-metal oxide composite material might be used as a chemiresistive sensor for breath-based disease detection (see here). Our efforts in this direction continue.

Organic semiconductors

The properties of organic semiconductors can also be fine-tuned through judicious molecular design. However, progress in this direction has been slow due to the exceedingly complicated relationship between single-molecule structure, crystal packing, and electronic properties in these materials. We entered this field by using a simple stochastic model to study how thermal molecular motion affects the localization of charge carriers in organic crystals (see here). More recently we have employed data science to design new organic semiconductors with targeted band gaps (see here) and showed how machine learning can be used to accelerate simulations of exciton diffusion in amorphous organic materials (see here). More is on the way!