DSpace@RPI (Rensselaer Polytechnic Institute)
Not a member yet
6809 research outputs found
Sort by
Incentivized Research Data Sharing, Reusing, and Repurposing with Blockchain Technologies
No full textData sharing is very important for accelerating scientific research, business innovations, and for informing individuals. Yet, concerns over data privacy, cost, and lack of secure data-sharing solutions have prevented data owners from sharing data. To overcome these issues, several research works have proposed blockchain-based data-sharing solutions for their ability to add transparency and control to the data-sharing process. Yet, while models for decentralized data sharing exist, how to incentivize these structures to enable data sharing at scale remains largely unexplored. In this paper, we study different incentive mechanisms for decentralized data-sharing platforms. Smart contracts are used to automate different payment options between data owners and data requesters. We evaluate multiple cost pricing scenarios for data monetization by simulating incentive mechanisms on a blockchain-based data-sharing platform. We show that a cost compensation model for the data owner rapidly cover the cost of data sharing and balance the overall incentives for all the actors in the platform
Designer DNA-based materials for DNA and virus sensing
No full textAugust 2021School of EngineeringOver the past few decades, the use of DNA has expanded from its well-known role as a carrier of genetic information to being exploited as a nanomaterial in its own right. Central to this exciting development is the programmable nature of Watson-Crick base pairing combined with its well-characterized and predictable nanoscale structure. Nano- and micron-scale structures built out of DNA can be combined with other functional materials like inorganic nanoparticles, biomolecules, and organic molecules to form unique hybrid structures with the combined functionality of both, the DNA and the functional materials. In this thesis, two novel functional hybrid DNA-based materials are investigated. First, the important and insufficiently understood biological phenomenon of Homologous Pairing (herein, HP) is implicated in the formation of phase-separated colloidal crystals upon annealing of binary mixtures of certain short DNA grafted polystyrene particles. The effect of DNA sequence and structure on the size of the colloidal crystals and the particle fraction that forms colloidal crystals is studied. DNA with higher GC content and isotropic bendability are found to be correlated with a greater extent of crystallization and therefore predicted to have a higher incidence of HP over their counterparts. The second DNA nanomaterial platform addresses a pressing societal need in the sensing of SARS-CoV-2, the virus that causes COVID-19. A tile-based 2-dimensional DNA nanostructure that spatially organizes multiple aptamers to target the spike receptor-binding domains (RBD) of SARS-CoV-2 was designed and synthesized. The DNA net-aptamer complex superimposes both the intra- and inter-cluster spatial patterns of the receptor-binding domains for maximum binding avidity. The SARS-CoV-2 spike RBD-specific aptamer was tagged with a fluorescent reporter, along with a quencher-labeled “lock” DNA that forms a partial duplex with the aptamer. The result is a highly sensitive and highly specific optical sensor that generates a fluorescence signal upon virus binding, with a limit of detection of 10^5 viral genome copies / mL in buffer and 10^3 viral genome copies /mL in the saliva matrix, and no cross-reactivity with related coronaviruses.Ph
Privacy preservation and evaluation in machine learning
No full textAugust 2022School of EngineeringThis thesis aims at improving methods for preserving privacy in machine learning for both generative and predictive models, and evaluating the privacy lost. Creating useful machine learning models requires the use of training data, but comes with a privacy risk to the subjects whose records are used. First, we address the problem of making sensitive data that must stay in a secured environment available for education or research by replacing it with sufficiently resemblant synthetic data that can be used to train accurate predictive models. Next, we consider the scenario in which researchers want to release models trained on sensitive data without the models revealing with high confidence which data were used for training. Two approaches are investigated and connections are made between them: making models differentially private, and protecting against membership inference attacks. We derive tight, componentwise bounds for the loss of a Wasserstein GAN, as well as new bounds on the norm of the loss of the gradient penalty term, and use those in a novel algorithm for a differentially private WGAN-GP. We evaluate the performance of this algorithm by using it to synthesize three real medical datasets, and using those synthetic datasets to replicate published medical studies. We find that the algorithm suffered less mode collapse than the non-differentially private version. We also develop a framework for formally analyzing the worst case privacy attack scenario. We prove several lower bounds on the accuracy of an attacker in this framework on model trainers that overfit or are insufficiently random. We also prove that any sample learned from incurs a risk of privacy loss, and that under certain assumptions black-box attacks are optimal. From our theoretical analyses, we motivate a novel protection method which we demonstrate can be used to improve the privacy of already well-protected models and simultaneously increase their accuracy.Ph
Applications of high throughput screening platforms for biologics discovery
No full textAugust 2021School of EngineeringNew drug discovery can be a very challenging and complex process due to the length and costs associated with the endeavor. To address this issue, high throughput screening (HTS) platforms have been instrumental in accelerating drug discovery by rapidly allowing us to test a large number of compounds simultaneously. This has been carried out in the past through the evaluation of small molecule drug candidates against relevant cell types in both two-dimensional (2D) and three-dimensional (3D) formats. More recently, novel biological therapeutics are increasingly being explored as treatments for several genetic diseases and cancer. Hence, it is important to develop similar HTS tools to study the production and evaluation of these biological products in the necessary context. In this thesis, the development of new HTS tools for studying such biological products are further explored. Multiple platforms are discussed, with each focused on different applications in viral vector process development, cancer immunotherapy, infectious disease research and stem cell differentiation. First, for viral vector process development, a high throughput 96-deep well plate platform was developed that enabled high density culture of HEK293 suspension cells for production of lentiviral (LVV) vectors. Using this platform, we were able to show that LVV production can be optimized in microscale cultures and that the results obtained were scalable to larger shake flask and bioreactor cultures. This system can serve as a valuable tool in early-stage bioprocess development. Second, for cancer immunotherapy discovery, a high throughput 330-micropillar microwell sandwich system was utilized to co-culture cancer spheroids with natural killer cells. Using this tool, we were able identify several natural killer cell-antibody-drug combinations that were most effective in causing cytotoxicity in different cancer cell lines. This information can help advance more personalized therapies for patient specific cancer. Third, for infectious disease research, we were able to develop a similar 532-micropillar microwell sandwich platform for screening new antivirals against SARS-CoV-2. We were able to show that a pseudoviral SARS-CoV-2 system can be used to successfully infect target cells on the platform. In the future, this system will be used to study different fucoidan compounds isolated from algae for their antiviral activity using the pseudoviral SARS-CoV-2 system. Finally, to study stem cell differentiation, we were able to establish a 384-pillar well sandwich platform to carry out CRISPR/Cas9 mediated editing in high throughput using an inverted GFP system on HEK293T cells. Moving forward, the roles of different genes in the differentiation process will be investigated in a rapid fashion by genetically modifying pluripotent stem cells with CRISPR/Cas9 and tracking that effect on the differentiation outcome.Ph
Fabrication of single to few layered graphene oxide membranes with tunable nanopores for separation of biomolecules
No full textDecember 2020School of EngineeringThe advent of 2 dimensional (2 D) materials opened new avenues in realizing an ideal membrane which has ultrathin thickness while not limiting the selectivity. Graphene oxide (GO) due to its water solubility and functionality has attracted a lot of attention for membrane application. However, the synthesis of an infallible and high performing membrane is still limited by thickness of GO membranes which leads to a loss of capacity of the membrane. In addition to this, tunable pores on GO is desired for its real-world application in various separation methodologies. Another challenge is developing single to few layered GO is the grain boundary defects which provide uncontrolled pathway for transport of molecules thereby limiting the selectivity. Alternative separation techniques are desired in the downstream of biotechnology industry due to the massive cost that they incur. Membrane technologies can offer a relatively non expensive alternative but the major problem with achieving the above targets is present use membrane with large pore size variation (log normal pore distribution of present polymeric membranes) will not give us adequate control over the process which is of prime importance as a regulatory standard as well as to produce desired product in biotechnology industry. Moreover, there is very limited literature on GO based membranes being used for protein separation.
Identifying the gap in literature gave us an impetus to work towards finding solutions for the aforementioned problems leading to amalgamation of 2 D material (GO) based membranes, and separations in biotechnology industry to outline my thesis. In this work, a new methodology (sequential deposition) is developed in order to fabricate single layered GO membrane with the vision of providing requisite selectivity without losing its inherent property of being one atom thick. With the objective of controlling and developing tunable transport pathways, oxygen plasma is employed for different intervals of time resulting in tunable pores on the surface of GO. We have also demonstrated rejection mechanism of different model proteins (BSA, Lysozyme & IgG) and mixed protein separation efficiency for similar sized molecules (Myoglobin and Lysozyme) with separation factor of ~6 & purity of 92% through the as synthesized GO membrane. These findings enhance the understanding of the synthesis of single layer GO membrane on polymeric support along with opening new avenues for fabrication of tunable GO membrane and its applications in biotechnology industry.M
Deciphering start dynamics in budding yeast using scanning number and brightness
No full textDecember 2021School of ScienceCell size homeostasis reflects the coordination between cell growth and division and is vital to all living organisms. This coordination is achieved primarily at the G1/S phase transition, termed Start in budding yeast. The major G1/S factors and their epistatic relationships are known. Cells pass Start upon activation of kinase Cdc28 by the G1 cyclin, Cln3, which phosphorylates the repressor Whi5, thereby activating expression of the downstream G1 cyclins CLN1/2 and ~200 other genes in the G1/S regulon by the SBF/MBF transcription factor complexes. It has been reported that dilution of Whi5 with cell growth controls the timing of Start while Swi4 concentration remains constant. Our previous results showed that the concentration of Whi5 remains constant with increasing size in G1 cells, while Swi4 is upregulated. In Chapter 4 of this thesis, single-cell imaging was used to quantify Whi5-GFP intensity as a function of time, rather than size, under experimental conditions that largely eliminate confounding effects of photo-bleaching. The results showed no significant time-dependent change in Whi5-GFP intensity in G1 cells. Measurements in a heterozygous WHI5 deletion diploid strain further validated the conclusion that Whi5 dosage is not a critical determinant of Start. In 2018, our group demonstrated that SBF/MBF subunit copy numbers, in particular Swi4, are sub-saturating with respect to their target promoters in small cells, but increase to a near 1:1 ratio as cells reach the Start threshold. That work also revealed that these factors, and consequently Cln1/2 levels, are upregulated in poor nutrients, conditions under which Start occurs at a smaller cell size. Chapter 5 of this thesis presents results describing a novel feedback loop for the expression of Swi4, the DNA-binding component of the dominant SBF transcription complex. It is likely that this additional regulatory mechanism contributes to the fine tuning of the Start transition. Finally, although it is known that the process of ribosome biogenesis is involved in regulating the cell size threshold, and moreover, is highly sensitive to external nutrient and stress conditions, the relationship between this pathway and the Start network remains to be determined. Preliminary results in Chapter 6 suggest the existence of a G1/S bypass mechanism that may come into play under conditions of decreased ribosome production. Future research will seek to further investigate this potential link between these growth and Start networks.Ph
Optimal allocation of parking spaces for heterogeneous vehicle types
No full textMay 2022School of EngineeringParking problems pose a huge burden to the economy, in the United States (US), drivers pay annually 72.7 billion is for cruising for parking, the rest of the cost is due to parking fines and overpaying. There are also major losses incurred by businesses, 39% of US drivers surveyed in 2016, reported avoiding shopping destinations where they know parking is limited, while 29% reported avoiding sports and leisure activities for the same reason. The main reasons behind these issues are limited availability of parking supply and lack of information about parking occupancies. This lack of information is the main reason forcing drivers to cruise for parking. Cruising has been estimated to be 30% of the traffic within cities, and it varies a lot between cities. In New York City, for example, it takes on average 15 minutes per trip to find a vacant on-street parking space, and 13 minutes per trip for off-street parking spaces.This research tackles these parking problems by building two optimization models that allocate individual vehicles (passenger cars, buses, delivery trucks, and service vehicles) arriving in a neighborhood to a specific on-street or off-street parking space, with the objective of reducing congestion, emissions and eliminate cruising. These two models include: (1) A static model in the case of small parking turnover, where the interarrival times are negligible; (2) A time-expanded model which considers variation in arrival times and the reusability of parking spaces. The proposed models take as inputs the vehicle and driver’s attributes from destination, parking duration, time of arrival and value of times (VOTs) along with information about the network and current parking occupancies. This information is used within an integer linear optimization problem to output specific destinations for individual arriving vehicles. Such models can serve as a core of a system maintained by cities to assign parking within smart cities. They can also be used by cities to optimally divide curbside parking by vehicle type and time of day. The results of the applied models show the superiority of the dynamic model, and it shows that vehicles with higher VOTs are better off parking near building entrances. The model also provides insights about scenarios of system breakdown that can be targeted by policy interventions.Ph
Metabolic Engineering of Saccharomyces cerevisiae for High-Level Production of Chlorogenic Acid from Glucose
No full textACS Synthetic Biology, 11, 800-811Note : if this item contains full text it may be a preprint, author manuscript, or a Gold OA copy that permits redistribution with a license such as CC BY. The final version is available through the publisher’s platform.Chlorogenic acid (CGA), a major dietary phenolic compound, has been increasingly used in the food and pharmaceutical industries because of its ready availability and extensive biological and pharmacological activities. Traditionally, extraction from plants has been the main approach for the commercial production of CGA. This study reports the first efficient microbial production of CGA by engineering the yeast, Saccharomyces cerevisiae, on a simple mineral medium. First, an optimized de novo biosynthetic pathway for CGA was reconstructed in S. cerevisiae from glucose with a CGA titer of 36.6 ± 2.4 mg/L. Then, a multimodule engineering strategy was employed to improve CGA production: (1) unlocking the shikimate pathway and optimizing carbon distribution; (2) optimizing the l-Phe branch and pathway balancing; and (3) increasing the copy number of CGA pathway genes. The combination of these interventions resulted in an about 6.4-fold improvement of CGA titer up to 234.8 ± 11.1 mg/L in shake flask cultures. CGA titers of 806.8 ± 1.7 mg/L were achieved in a 1 L fed-batch fermenter. This study opens a route to effectively produce CGA from glucose in S. cerevisiae and establishes a platform for the biosynthesis of CGA-derived value-added metabolites.National Natural Science Foundation of Chinahttps://login.libproxy.rpi.edu/login?url=https://doi.org/10.1021/acssynbio.1c0048
Data-driven strategies for control in additive manufacturing
No full textDecember 2021School of EngineeringAdditive Manufacturing (AM) techniques are quickly becoming attractive for fabricating parts ranging from biological tissues to aircraft components. A key challenge in these techniques is controlling the quality of parts to ensure high accuracy, throughput and repeatability. Given the fast and high-dimensional multi-scale dynamics associated with many AM processes, it is often difficult to physically model these processes for control. This thesis examines how data obtained from the process may be used to implement feed-forward and feedback control. Two AM processes are studied: Inkjet 3D Printing, commonly used for fabricating polymer parts; and Selective Laser Melting (SLM) used for producing metal parts. Inkjet 3D printing builds, or prints, 3D parts with high precision by sequentially depositing and hardening liquid material. Because of the complex fluid dynamics at play and multiple printing parameters to tune, the dynamic process is challenging to model. There is significant prior work on modeling local droplet behavior, without necessarily considering the overall part-level dynamics. On the other hand, studies concerned with part geometry control have developed reduced order models that do not well capture nonlinear fluid behavior. In this research, to implicitly learn the fluid dynamics for geometry-level control, we employ machine learning strategies but push their conventional usage in the following ways: (1) We use a shallow multi-layer neural network to capture the geometrical relationship between printing input pattern and measured output profiles. (2) We develop a data-driven physics-guided model that uses a recurrent neural network to model droplet spreading and coalescence. Not only does this model capture the complex fluid behavior, it is formulated to do so at the geometry level. We validate the model on data collected from an inkjet 3D printing setup. The proposed model outperforms a blackbox off-the-shelf multilayer perceptron (neural network) by using significantly less data for training, at the same time delivering better performance in RMS error on test data. The proposed model is also compared with a state-of-the-art reduced order linear model and shows substantial improvement in RMS error on test data. Experimental results also underline that the model parameters learned are geometry invariant, that is, the model parameters trained on one geometry can be used to predict the height map evolution for other geometries without relearning. Next, we propose and demonstrate a new predictive control algorithm that leverages the neural-network-like structure of the model. Back-propagation is used for efficient gradient calculations to determine optimal control inputs, namely droplet patterns for subsequent layer(s), to optimize a quadratic cost function. Further, we analyse the stability of the open-loop and closed-loop printing system based on the developed model and control scheme. In addition, we develop an efficient algorithm to make online learning and feedback control practically feasible purposes. Simulation and experimental results show that both feedforward and feedback control substantially improve the height profile over existing control approaches. In line with the data-driven theme of this thesis, we investigate a data-driven approach for controlling a metal AM process, selective laser melting (SLM), where a laser beam is to local melt metal powder and create complex geometries in a layer-by-layer manner. Similar to the inkjet process, because of the complexity associated with the heating, melting, cooling and solidification, it is difficult to model the SLM process in a control-oriented fashion. To address this, this thesis presents a model-free iterative learning control (ILC) scheme for designing laser power profiles for multi-objective temperature control in SLM. The goal is to ensure that while temperature distribution in the selected region is sufficient to cause melting and fusion, the meltpool is not overheated. We first formulate this goal as an optimization problem with the power profile as the decision variable and the cost function to be minimized being the sum of two unidirectional error terms (for upper and lower temperature bounds, respectively). Given the difficulty in analytically modeling the temperature-laser power relationship in SLM for gradient computations as in standard ILC, we solve the minimization problem using a model-free ILC scheme. In this scheme, the control input that minimizes the cost function is learned through a data-driven gradient descent update that uses the process itself to compute the gradient direction. The gradient descent algorithm proposed here accounts for the time-varying behavior of the SLM thermal dynamics because of the scan path. This is accomplished by feeding the temperature output error, reversed in time, through the process itself with a reversed scan path direction. For validation, this multi-objective gradient-based ILC algorithm is implemented on a three-phase high-fidelity simulation of the SLM process. The results demonstrate the algorithm's ability to drive the temperature distribution to within a prescribed range in scenarios where standard (single-objective constant gain) ILC fails.Ph
The impact of surface modification on magneto-functional iron oxide – polyethylene oxide nanocomposites
No full textAugust 2022School of EngineeringRemote triggering of smart materials such as shape memory polymers and nanocomposites for drug delivery are research areas of continued interest for magneto-functional nanocomposites. Magnetically susceptible nanoparticles (NPs) generate heat when exposed to an alternating magnetic field (AMF), making these NPs an ideal candidate for use in smart nanocomposites. While a wide range of nanoparticle chemistries have been studied as ferrofluids in various liquid carrier media, the behavior of these nanoparticles in solid state polymers is not widely understood. This research studies the dependence of heat generation mechanisms and interfacial interactions on nanocomposite structure and morphology, nanoparticle surface coating and nanoparticle concentration in iron oxide (Fe3O4) - poly(ethylene oxide), PEO, nanocomposites. Nanoparticles were coated with surfactants and polymers to improve dispersion and magnetic properties. In this work we first focused on the impact of surface coating of iron oxide (Fe3O4) NPs on magnetic volume reduction, structure, magnetic heating efficiency and mechanical properties of poly(ethylene oxide), PEO, nanocomposites. Uncoated, poly(ethylene glycol), PEG, coated and amine coated 10–nm–diameter Fe3O4 NPs were dispersed at concentrations less than 1% by weight in PEO. Although loaded at low concentration these nanocomposites displayed excellent values for intrinsic power loss especially at low concentrations. We found that dispersion of nanoparticles was strongly related to the character of the surface coating. Uncoated nanoparticles formed large aggregates which led to a significant decrease in the heat generation capabilities. The surface coatings also strongly impacted the magnetic phase reduction. Amine coated nanoparticles had the least magnetic phase reduction. All nanoparticles showed unexpectedly higher heating efficiencies in PEO than when dispersed in water due to decreased magnetic volume loss. Aggregation was determined to be the dominant factor for decreased heating efficiency. Calorimetry experiments explored the impact of the nanoparticles on crystallinity and nucleation rates. Nanoindentation was used to evaluate the mechanical properties via stress relaxation and creep experiments. Amine coated nanoparticles were found to improve the moduli of the nanocomposites. Low concentrations of nanoparticles led to increased relaxation and decreased creep compliance whereas high concentrations had no effect on relaxation and increased creep compliance. The relevance of five rheological models was evaluated. Stress relaxation was best modeled by a power law or logarithmic based model whereas the creep was best modeled by a Generalized Maxwell model.
In the second part of this work, single core aminosilane coated 10–nm–diameter Fe3O4 NPs were dispersed at concentrations less than 2% by weight in PEO matrices with varying molecular weights. Altering the matrix molecular weight of the matrix polymer allows for consistent intermolecular interactions between the NP surface groups and the PEO in order to determine relative importance of Brownian and Neel relaxation processes. Increased matrix molecular weight above the polymer matrix entanglement molecular weight led to decreased heat generation efficiency that was consistent with decreases in the nanoparticle magnetic volume determined via vibrating sample magnetometry. Brownian and Neelian relaxation mechanisms were proven to be present despite the high viscosity of the matrix media. Dynamic polymer relaxation modes such as the Reptation or Rouse models were found to be inactive.Ph