Performing computational molecular dynamics (MD) simulations of small-molecule systems has become one of the most prominently used methods in studies of molecular structure and behavior. However, MD is a computationally expensive and time-consuming methodology because of the requirement of computing detailed interactions among atom-atom pairs. There is great interest, therefore, in reducing the time and computational power needed to approximate real-world systems. Most commonly, such efforts have employed machine learning techniques to predict extensive properties of molecular systems. Here, we propose accelerating simulations by predicting conformational changes - a prospect that has not yet been fully explored. Previous work attempted applying a linear dynamical analysis algorithm named Dynamic Mode Decomposition to MD data, which has been shown to be ineffective through a multiresolution analysis. We propose herein the use of Sparse Identification of Nonlinear Dynamical Systems (SINDy), a nonlinear model which has been shown to accurately decipher the governing equations of dynamical systems. We will be testing the effectiveness of SINDy with MD data by performing an iterative error analysis while varying the initial parameters of the dataset, thereby gaining a better understanding of how much data (and in what form) should be inputted to maximize the accuracy of a simulated SINDy model of an MD dataset. If shown to be sufficiently accurate, we then can implement SINDy simultaneously with MD in an active learning loop to save time and computational power while maintaining a high degree of predictive capability for peptide conformations. The current goal is to obtain a deeper understanding of peptide conformational changes that could, in the future, be combined with machine learning techniques to greatly accelerate classical MD simulations.

This project is supported by the UW Computational Neuroscience Center, and the DMREF Program of NSF through the MGI platform under DMR# 1629071, 1848911, and 1922020.

The goal of this project is to develop a dynamically evolving connectionist model that more closely resembles the brain through its information-processing. Over the years, AI has shifted from the first generation of feedforward systems to the use of recurrent or convolutional Neural Networks. The third and newest generation of AI models, the brain-based models, and the Spiking Neural Network (SNN), attempts to bridge the gap between Neuroscience and ML using biologically realistic models like Θ-model, LIF, Izhikevich, HR, HH. These models, however, are still a black box leaving very little control or understanding on the learning process within the system without the access to the inner structure of the network. In addition, these systems are highly inefficient, slow, and very complex due to the limitations imposed by the hardware and explicit simulation of partial differential equations. Real world problems require “flexible learning and dynamically adaptive connectionist systems” that are capable to adapt and accommodate new input in real time. Current solutions have focused on varying the weights within a system rather than focusing on how connections within the system are formed. Based on our understanding from organismal brain structures, our approach, called biomimetic information codec, .bic, is a morphologically-adaptive coding hierarchical network that form in accordance with energy minimization - driven by dissipation of "heat" generated by the training data - constructing cortices and connectome for processing of information. My first objective herein is to quantitatively compare detailed structures between biological (fly brain) and .bic. networks using a random matrix approach.

Loss of periodontal ligament tissue (PDL) and attachment is a serious complication of periodontal diseases - the most prevalent dental health problems. PDL-degeneration leads to alveolar bone degeneration, infection, gingivitis, and eventual tooth loss. There is currently no product that can cure PDL-degeneration as regeneration requires the combinatorial process of regenerating cementum, signaling the existing relevant cells to proliferate and form PDL, and its integration into a functional system. Current restorative treatments utilize cell-based tissue regeneration, synthetic scaffolds, tissue grafts with limited, temporary success. A market product, e.g., claims to restore periodontium using harvested fetal swine periodontal tissue with highly variable clinical outcomes. Although these traditional procedures are well-established and show some success, their efficacy is limited due to the lack of structural and functional integration of a deposited layer with the underlying tooth, specifically integration into the remineralized cementomimetic layer. GEMSEC labs have developed a proprietary technology dubbed “peptide-guided remineralization” which facilitates new mineral formation using protein-derived peptides and have successfully restored dental hard tissues via several case studies including enamel, cementum, dentin under in-vitro and in-vivo conditions. Translating this technology into a daily-use product, we propose a PDL-regenerating chimeric construct which includes a biomineralizing peptide, ADP5, derived from the key enamel protein, amelogenin, with cell signaling moieties. Herein, we aim to use established bioinformatics, machine-learning tools, and high-throughput experimentation to identify peptides from proteins involved in PDL development cell-signaling towards controlled biomineralization, bioadhesion, and cell-signaling functionalities necessary for PDL regeneration. Addressing current treatment protocol limitations, the interdisciplinary approaches developed in this project are designed for the regeneration and formation of fully functional PDL. 

Engineered solid binding peptides can be used as molecular tools for a variety of bio/nanotechnology applications, especially in interfacing biology with solid-state devices at bio/nano soft interfaces. The control of surface organization, and therefore peptide-solid interactions, is critical and involves surface phenomena such as binding, surface diffusion, and self-organization on atomically flat solids. Each of these phenomena requires the knowledge of peptide’s folding patterns which are, however, difficult to study both experimentally and computationally. Molecular dynamics, MD, has been used to computationally model peptide/solid interactions, but without information regarding the energy landscape of peptide conformations the challenge of predictive design remains. While several methods exist for finding the energy landscapes of single peptide systems, currently no approach handles multi-peptide/surface systems. Here we use Time-Varying Autoregression with Low Rank Tensors, TVART, to efficiently explore the energy landscapes of such systems, aiming to find accurate linear approximations for predictive design of peptides at bio/nano interfaces. Using TVART, with each slice representing a discrete time window, allows for temporal smoothness and high predictive accuracy. It is anticipated that some descriptions of conformation will be better suited to describe peptide conformation energy landscapes than others; based on this premise, we examined interatomic distances/adjacencies and peptide backbone torsion angles as descriptions of peptide conformation. Through such analyses, it is becoming possible to describe how peptide conformations in multi-peptide/surface systems evolve through the energy landscape and settle into energy minima (stable conformations). These conformations can then be corroborated with experimental validation of peptide self-organization on the surface using scanning probe microscopy techniques with sub-A resolutions. The combination of computational modeling and high-resolution experiments is expected to aid predictive design platforms for future applications in biosensors, bioelectronics, and logic devices.

Glioblastoma (GBM) is a cancer originating in glial cells in the brain that accounts for more than 60% of all brain tumors in adults. The low survival rate can be attributed to high resistance to radiotherapy due to the hypoxic tumor environment which induces signaling networks in cancer cells that lead to the epithelial to mesenchymal transition (EMT). EMT gives rise to mesenchymal cancer stem cells (MSC) with a highly invasive phenotype which resists traditional means of therapy. Phospholipid glutathione peroxidase (GPX4), a selenocysteine-containing enzyme that dissipates lipid peroxides, has been shown to regulate pathways that prevent ferroptosis, a unique iron dependent form of cell death initiated by an increase in reactive oxygen species. Disrupting the GPX4 pathway by siRNA-induced gene knockdown induces ferroptosis. Therefore, NPs as a vector for gene therapy may be able to eliminate mesenchymal state stem cells for a more effective treatment. Using hypoxia to induce EMT to develop a cell model for this work, preliminary results from quantitative real time PCR showed a correlation between GPX4 and EMT markers of human glioblastoma cells in hypoxia. GPX4 siRNA were evaluated using commercially available transfection agents on hypoxic and normoxic cells as proof of concept in vitro over a period of ten days. NP mediated delivery of validated siRNA were optimized using different ratios of NP and siRNA. Incubation time was also optimized. Finally, dual therapy of siRNA knockdown and radiotherapy were performed to evaluate sensitization of cells. The capabilities of NPs, along with concurrent radiation therapy, may provide a means to overcome radioresistance in GBM therapy.

Dental implantation is a common clinical procedure used to replace missing teeth and maintain bone structure and facial aesthetics. However, it leads to unexpected side effects, including bone loss or peri-implantitis in 1 out of 10 cases due to failure of osseointegration, defined as improper integration of the implant into the mineralized bone. To enhance osseointegration, the present study aimed to form a layer of hydroxyapatite that could facilitate the integration of the implant (Titanium or Zirconia) with the alveolar bone that lead, while also having antimicrobial property to prevent local infection. Our previous study demonstrated that titanium-binding peptides (TiBPs) are able to bind specifically to the surface of Ti and that amelogenin-derived peptide (ADPs) can be used for direct remineralization on the bone surface. We also identified antimicrobial peptides (AMPs) that can inhibit common oral bacterial growth. These results imply that there is an opportunity to design two of heterofunctional peptides, both binding to Ti with one has the function of directing biomimetic remineralization process, while the other providing antimicrobial activity. The Ti-surface is modified by chimerizing the TiBPs and the ADPs, and TiBPs and AMPs, with short amino acid sequences. The overall process is separated into two main steps: 1. Designing and synthesizing the chimeric peptide; 2. Characterizing (a) binding, (b) mineralization and (c) antimicrobial efficacy of chimeric peptides on the implant surface. We predict that the chimeric peptides will have high binding affinity to the titanium surface while simultaneously enabling mineralization on the implant surface and inhibiting the growth of bacteria. The present aims to contribute to the foundation of finding a long-term novel dental implant treatment via the molecular biomimetic approach towards a clinical strategy to enhance the long-term durability of dental implants. The research is supported by Mary Gates Scholarship (YL), Spencer Funds from School of Dentistry, and CoMotion.

Single amino acid conformational preferences on a substrate are invaluable to our understanding of how conformational propensities are dictated by a peptide’s sequence. The efficiency of understanding SAP/SLAM (self-assembling peptides on single layer atomic materials) interfaces is of high importance. An efficient way of performing computational modeling of a peptide’s free energy landscapes is needed to predict the folded structures on solid surfaces towards designing bio/nano interfaces, a key to bioelectronics and biosensor developments. Graphene, single-atomic layer graphite, is an ideal substrate for a dodecapeptide to bind and spontaneously self-organize to form ordered biomolecular structures on the surface. For the purpose of shortening computation times of peptides at graphene interfaces, sampling each amino acid’s free energy landscape in terms of the peptide’s natural torsional configuration, such as the phi and psi angles of the peptide backbone on graphene is studied. This is due to the fact that a predetermined natural starting point, such as the peptide’s lowest energy structure will allow for quicker convergence of the system and more accurate structure prediction. In an effort to retrieve this data, here we simulate each amino acid using the enhanced sampling computational technique Metadynamics. Using the computational modeling and random sampling of the lowest energy wells, we aim to aid in the determination of low energy preferences on conformational landscapes on graphene towards more predictive design of soft bio/nano interfaces for practical implementations. 

In this project, we measure electron flow in graphene, a 2-D lattice of carbon atoms, and compare the results to simulations that we run. As current is passed through typical electrical devices, electron transport is dominated by momentum relaxing electron-phonon scattering, i.e. electrons colliding with the impurities and vibrations of the crystal's lattice structure. This is typical of the omhic regime. However, other modes of electron transport are possible. In clean graphene, for example, electrons are weakly coupled to lattice sites and electron-electron scattering dominate. In these interactions, momentum transferred between electrons is conserved. When measured over a range of temperatures, we find dips in the resistance, resulting from these hydrodynamic electrons’ tendency to “pull” one another along with the bulk. Analogous to honey, these electrons have viscosity, which unlike resistivity, is a property of the fluid. This research will further elucidate properties of this electron fluid. To complete this project, we will fabricate graphene devices and study them in a table-top cryostat, measuring the current output from 4K to room temperature. We are particularly interested in how this viscous fluid behaves as it encounters a boundary within the device, an open question in the field of solid-state physics. We use a low voltage probe tip which can be positioned anywhere within the device. By blocking a portion of the drain with the probe-tip and measuring the current output along segments of the drain, we may gain insight into the boundary conditions of the electron fluid. This research will directly benefit the electronics industry: the next generation of computer chips will utilize 2-D materials such as graphene, potentially enabling the useful properties of hydrodynamic flow to be exploited.

The purpose of my research is to explore the protein-ligand binding interaction by using sequence encoding and signal analysis to process and understand amino-acid sequences. Applications of this technique would be useful in a variety of biomedical fields, but more specifically in creating a platform for a unique and streamlined vaccine candidacy process. If we can encode protein/DNA sequences using electron-ion interaction-potential, then by applying various signal processing functions we can more easily identify the meaning of these complex structures. The goal in doing this would be to relate simple amino-acids, the building blocks of our genetic sequences, to numerical values which we may manipulate. Combining these inferences with a convolutional neural network, the result would create a non-empirical and efficient method of understanding protein folding and bimolecular binding interaction, specifically through predicting pIC50 binding affinity. The methodology begins with accessing a large set of human amino-acid sequence data, transforming the data using signal processing, and learning from the data to understand similarities between sequences and predict binding affinity. The anticipated results will initially be the neural networks ability to predict binding affinity. Further results and experimentation would involve tuning the model to be more adaptive and testing new data from the SARS-Cov-2 virus. The more variety of human data the model may learn from, the more adaptive and accurate it will predict pIC50 values in an efficient, non-empirical manner.