(4hi) AI-Powered Protein Engineering for Clean Energy and Biomedical Applications

Apply and develop molecular simulation, statistical thermodynamics and machine learning methods to 1) investigate the self-assembly and phase transition of soft matter and biomolecule solutions; 2) engineer proteins for clean energy applications such rare earth separation and biomedical applications such as biomaterial, antimicrobial therapeutics, lipid nanoparticle based drug delivery.

Teaching Interests

With holding both bachelor and PhD degrees in Chemical Engineering, I am confident to teach any core chemical engineering courses, at both undergraduate and graduate level. In addition, I am particularly interested in teaching courses including molecular modeling, statistical mechanics, machine learning, molecular engineering and design, biophysics, and interfacial phenomena.

Abstract

Proteins are one of the most common and essential building blocks for living systems. Proteins have become excellent candidates for various engineering applications due to their rich structural and physicochemical properties arising from large sequence spaces. Here, I demonstrate how a chemical engineer integrates modern molecular modeling and machine learning methods to understand and quantify various protein functions for clean energy and biomedical applications.

In the first project, I elucidated the self-assembling thermodynamics of Amyloid β peptide by computing the first experimentally-verified Aβ_16-22 peptide solubility phase diagram. The work demonstrates the power of integrating machine learning, molecular modeling and nucleation theory in resolving amyloid formation mechanism related to the Alzheimer’s disease pathology. Then I translated the knowledge of amyloid formation into designing co-assembled peptides for functional nanofiber fabrication.

In the second project, rare earth elements (REE) are critical materials with numerous clean energy applications including batteries, semiconductors, and electric vehicles, due to their unique luminescent, magnetic, and catalytic properties. A bio-inspired REE separation scheme was developed by engineering the calcium-binding motif of bacterial proteins into the design of selectively lanthanide-binding peptides. I first performed enhanced sampling simulations to reveal the molecular binding mechanism of peptides and trivalent REE cations in the aqueous solution. Then I employed physics-driven machine learning algorithms to predict peptide-REE binding affinity and selectivity, paving the road for developing green and efficient alternative rare earth extraction and recovery.

In the third project, kinases are protein enzymes that catalyzing post-translational phosphorylation and are essential for cell regulation and hemostasis. The extent to which kinases are activated due to mutations specific to each cancer patient can profoundly impact disease progression and drug efficacy in clinical trials. Consequently, pinpointing the impact of somatic mutations in the human kinome in their functionality emerges as a critical step towards personalized cancer therapy. I developed a machine learning algorithm (Cancer AI) to predict the impact of mutation on the catalytic activity of cancer-driving kinase protein enzymes. The Cancer AI tool is incorporated into the clinical treatment of cancer (neuroblastoma) patients led by oncologists.