Selective targeting of damaged or diseased cells is a concept with great potential to revolutionize the efficacy of systemically administered agents. Successful targeting preferentially delivers imaging agents and/or therapies to specific tissues, enhancing detection and diagnosis as well as therapies minimizing damage to nontarget cells. However, the current generation of targeted therapies has not yet generally achieved highly specific targeting of tumors and emerging neoplasia through a single recognition mechanism. Effective performance of targeted systems depends upon recognition of unique characteristics on the cellular surface. HT-1080 cells in vitro present 3,840,000 ± 70,000 CD13 receptors per cell, a level presumably suitable for effective targeting. Receptor presentation can be further modulated by exposure to factors such as ionizing radiation and various cytokines, presenting opportunities to modify receptor expression for optimized delivery of targeted constructs. Single modality targeted nanoscale quantum dots (QDs) were synthesized to enable specific interaction with target cells. Nonspecific interaction was limited by functionalizing the QD surface with a passive PEG coating. To facilitate specific binding to target receptors, QDs were surface-functionalized with targeting peptides. The resulting constructs (QD-PEG-NGR) bound to the surface of HT-1080 cells at a level of 15,410 ± 980/cell, enabling high contrast imaging and the potential for significant therapeutic delivery. To overcome specific limitations that plague current single-modality targeting technologies, a multifunctional QD-based proximity-activated (PA) targeting system was developed. QDs were functionalized with a proteolytically sensitive PA coating designed to mask an underlying targeting ligand. Specific matrix metalloprotease-7 (MMP-7) activity resulted in maximal cleavage of 90.9 ± 15.4% of the available PA structures from the QD surface. Effective cleavage was measured at exposure times and enzyme concentrations consistent with estimated in vivo conditions. Multifunctional NPs offer opportunities unavailable with molecular structures or current single-modality targeted constructs. Effective targeting, imaging, and delivery to specific cells can be achieved with the two-component targeting methodology detailed in this work. Application of these targeting methodologies may offer a powerful system to significantly improve cancer treatment.

Fundamental clinical approaches for assessing patient vital signs have changed little since the first invasive blood pressure measurements were made over 100 years ago. Interpreting patient physiology remains largely a manual, intermittent process, despite evidence suggesting that automated processing of continuously-captured physiologic data will yield new, important measurements. These new vital signs may predict patient improvement or deterioration, and signal specific opportunities for early therapeutic intervention in clinically meaningful, cost-effective ways. However, tools and methods to discover, refine, and validate new vital signs in working clinical settings, across large patient populations, have been lacking. This work describes the SIMON (Signal Interpretation and Monitoring) system, and its application to the discovery, refinement, and validation of a prototype new vital sign, integer heart rate variability (HRV). SIMONs modular architecture enables a high degree of reliability and scalability for dense physiologic data capture, processing, and decision support tasks. The system has been in use continuously since 1998 in the Vanderbilt trauma intensive care unit (ICU), provides physiologic data reporting, display, and alerting capabilities, and has archived physiologic data from over 3500 patients. Its alphanumeric pager alerting functionality has been evaluated in the domain of intracranial pressure management. Additionally, a new measurement of HRV has been developed, refined, and validated in a population of over 1000 trauma patients. The result is not only a new predictor of mortality but also represents proof of concept that a working intensive care unit can serve as a rich, automatic source of data to discover new predictive patterns in patient physiology. Ultimately, study of HRV and other new vital signs may correlate failure of the autonomic nervous system or other neural and hormonal communication pathways with specific injuries, diseases, or patient characteristics. These studies could, in turn, illuminate regulatory mechanisms uniting systems, organs, cells, proteins, and genes. Such knowledge provides a basis for additional research, and informs design of the next generation of ICU monitors and decision support tools to improve quality and efficiency of medical care.

This thesis describes the development and evaluation of a computerized algorithm for detecting patients with acute asthma exacerbations who present to a pediatric emergency department (ED). A rule-based algorithm was designed to collect patient information from the computerized patient record at the time of ED triage. We confirmed the feasibility of this approach through a retrospective analysis. The algorithm was then implemented in the pediatric ED as a real-time asthma detection system. Its performance was evaluated prospectively during a two-month study period on over 3,500 ED patients, of which 342 had an asthma exacerbation. The system was able detect patients presenting with acute asthma with high accuracy. Sensitivity was 71.6%, specificity was 97.8%, positive predictive value was 77.0%, and negative predictive value was 97.1%. This research could be applied to detect and automatically initiate guidelines for the management of asthma in eligible patients, and could serve as a model for detecting other conditions which are managed by standardized guidelines in the ED.

Recent studies have demonstrated that organotypic raft cultures serve as an excellent model for the optical behavior of actual tissues. Other studies in our lab have demonstrated that Raman spectroscopy can discriminate among basal cell carcinoma, squamous cell carcinoma (SCC), non-normal benign, and normal skin in vivo. The primary purpose of this work is to understand the biologic basis of skin Raman spectra and to determine the impact of the location of SCC cells within raft cultures on the spectral bands. Rafts were constructed with SCC cells in the stroma or in the epidermis, and measurements were taken at multiple time points with a Raman probe system. The data allowed tissue discrimination as before, and they showed that the location and concentration of SCC cells do not have a large influence on macroscopically-gathered data. In addition, data gathered from the epidermal vs. stromal layers, combined with histology, support the hypothesis that Raman spectroscopy can detect biochemical changes associated with malignancy before such changes are evident via histology.

Coronary heart disease is the single largest cause of mortality in the United States. Approximately 335,000 people die annually from sudden cardiac death, and the majority of these cases are believed to be from ventricular fibrillation. To effectively treat and prevent cardiac rhythm disturbances, the response of the heart to electrical stimulation must be understood. Although cardiac defibrillation therapy is an invaluable medical procedure, the mechanisms by which strong electrical shocks terminate potentially lethal fibrillation are still debated. Bidomain models of cardiac tissue successfully characterize many of the effects of electrical stimulation of the heart. In the bidomain, the intracellular and extracellular spaces are distinct and have differing electrical anisotropies. With unequal anisotropy ratios, bidomain theory predicts simultaneous positive and negative polarization in response to stimulation, in the form of virtual cathodes and anodes that lead to interesting cardiac activation dynamics. This research examined experimentally the response of cardiac tissue to electrical stimulation from a bidomain perspective. Changes in transmembrane potential during and following electrical stimulation were recorded optically using a voltage-sensitive fluorescent dye. Optical mapping allows noninvasive measurement with high spatiotemporal resolution and avoids electrical stimulus artifacts. We found that: 1. During unipolar anodal stimulation of diastolic tissue, the mechanism of excitation depends upon the extracellular potassium concentration. 2. With proper timing of unipolar stimulation close to refractoriness, damped waves with diminished amplitude and velocity either gradually die or sharply increase in amplitude after a delay to become a steadily propagating wave. 3. We confirmed bidomain model predictions that virtual electrodes from unipolar stimulation affect excitability through the cardiac cycle as shown by strength-interval curves. 4. Field stimulation of the diastolic heart revealed that increasing shock strength and duration do not necessarily result in faster activation because of virtual anode polarization. 5. Alternating regions of positive and negative virtual electrode polarization around an insulating heterogeneity occur during field stimulation and may affect plunge electrode measurements. An increased understanding of how cardiac tissue responds to electrical stimulation in various conditions will guide improvements in treatment and prevention of cardiac rhythm disorders.

Two attractive features of ELISA are the specificity of antibody-antigen recognition and the sensitivity achieved by enzymatic amplification. We describe the development of a non-enzymatic virus detection platform based on circumferential bands of antibody probes coupled to a 120 mm diameter polyester filament. Automated processing was achieved by sequential positioning of filament-coupled probes through a series of liquid filled glass microreaction chambers. Antibody regions were first positioned within a microcapillary tube containing a solution of M13KO7 virus before being moved through subsequent chambers, where the filament-coupled probes were washed, exposed to a fluorescently labeled detecting antibody, and washed again. Using anti-M13KO7 mAb coupled to a polyester filament, the presence of 8.3 x 108 M13KO7 virus particles produced a 30-fold increase in fluorescence over an immobilized negative control antibody. Similar to ELISA, this filament-based approach had a lower limit of sensitivity of ~1.7 x 107 virus particles. We then combined the automated filament processing with an integrated laser-based optical detector to enable real-time controlled detection of virions in solution. A 638 nm laser with a photomultiplier at a right angle provided continuous monitoring for the presence of the fluorescently labeled detecting antibody. A virus incubation time of 1 minute detected 1010 virions/mL. Repeated incubations of antibody regions in either the virus or labeled antibody chambers increased fluorescence roughly proportional to the incubation times. This technology was used to identify and characterize a reovirus strain. We developed a decision tree that tested for reovirus with increasing specificity at each level of the tree. Using three types of reovirus and one bacteriophage, our system correctly detected and identified all three reovirus strains at a concentration of 2 x 1012 virions/ml and M13K07 phage at 3 x 1011 virions/ml. Fluorescence from all peak regions was determined to be significantly higher that background regions (p < 0.05). Using online feedback to guide testing, this scheme could easily be expanded into a much more complicated system with numerous levels and branches. This platform may prove attractive for point-of-care settings, the detection of biohazardous materials, or other applications where sensitive, rapid, and automated molecular recognition is desired.

Unwanted drug-drug interactions endanger millions of patients each year and burden families and the hospital system with escalating costs. Computer-based alerting systems are designed to prevent these interactions, yet the knowledge bases that support these systems often contain incomplete, clinically insignificant, and inaccurate drug information that can contribute to false alerts and wasted time. It may be possible to improve the content of these drug interaction databases by facilitating access to new or underused sources of drug-drug interaction information. The National Library of Medicine's MEDLINE database represents a respected source of peer-reviewed biomedical citations that would serve as a valuable source of information if the relevant articles could be pinpointed effectively and efficiently. This research compared the classification capabilities of human-generated and computer-generated Boolean queries as methods for locating articles about drug interactions. Two manual queries were assembled by medical librarians specializing in MEDLINE searches, and three computer-based queries were developed using a decision tree modeled on Support Vector Machine output. All five queries were tested on a corpus of manually-labeled positive and negative drug-drug interaction citations. Overall, the study showed that computer-generated queries derived from automated classification techniques have the potential to perform at least as well as manual queries in identifying drug-drug interaction articles in MEDLINE.

In light of the growing scarcity of livers available for transplantation, careful decisions must be made in organ allocation. The current standard of care for transplant decision making is the use of clinical judgment, although a good model to predict survival after liver transplantation may be useful to support these difficult decisions. This thesis explores the use of informatics techniques to improve upon past research in modeling liver transplant survival. A systematic literature revealed that the use of machine learning techniques has not been thoroughly explored in the field. Several experiments examined different modeling techniques using a database from the United Network for Organ Sharing. A Bayesian network was created to predict survival after liver transplantation, and it exceeded the performance of other models published in the literature. Fully automated feature selection techniques were used to identify the key predictors of liver transplant survival in a large database. A support vector machine was used to show that a relatively simple model, consisting of main effects and two-way interactions, may be adequate for predicting liver transplant survival. A pilot study was conducted to assess the ability of expert clinicians in predicting survival, and they tended to perform similarly to mathematical models. The results lay a foundation for future refinements in survival modeling and for a clinical trial of decision support in liver transplantation.

A persistent limitation to current and future anti-cancer therapies is an inability to deliver high levels of active agent throughout the entire target tumor tissue volume without adversely impacting normal tissues. Interstitial transport is compromised by the poor mobility of macromolecules and nanoscale structures. We developed an in vitro system to quantify the facilitated transport of superparamagnetic (SPM) nanoparticles (NPs) through model extracellular matrix (ECM) in vitro. SPM NP motion was induced by an external magnetic field. 135 nm radius NPs with a polyethylene glycol (PEG) surface moved through the ECM with an average velocity of 1.5 mm h-1, a velocity suitable for some clinical applications. Steric barriers such as collagen I sharply limit interstitial delivery of macromolecular and nanoparticle-based therapeutic agents. Collagenase-linked SPM NPs overcame these barriers and moved through ECM in vitro at 90 ìm hr-1, a rate similar to invasive cells, under the influence of a magnetic field. Temporal decay of collagenase activity shifted from an exponential behavior in solution to a linear relationship when NP-attached. NP platforms offer the opportunity to develop a unified synthesis method for formation of multifunctional agents. We have demonstrated a facile method of conjugating multiple enzyme species to a NP via sulfhydryl-maleimide reaction chemistry. Horseradish peroxidase, á-glucosidase, and collagenase were simultaneously conjugated to the 300 Da PEG surface of SPM NPs at 15:1, 127:1 and 103:1 functional enzyme:NP ratio, respectively. SATP addition of sulfhydryl groups to each enzyme was achieved without significant reduction in enzyme activity. Cross reactivity of enzymes between enzyme activity assay systems was negligible. Multifunctional NPs mimic complex invasive biological processes found in metastatic invasion and immune cell interactions. Isolated study of sets of enzymes in an invasion experimental system may make an ideal screening tool for blocking pathogenic invasive processes, including cancer metastasis and abnormal angiogenesis. Dispersion of otherwise immobile macromolecular or nanoscale therapeutic structures can be achieved with the proteolytic SPM NP carriers detailed in this work.

In biomedical and biological domains, researchers typically study continuous data sets. In these domains, an increasingly popular tool for understanding the relationship between variables is Bayesian network structure learning. There are three methods for learning Bayesian network structure from continuous data. The most popular approach is discretizing the data prior to structure learning. Alternative approaches are integrating discretization with structure learning as well as learning directly with continuous data. It is not known which method is best since there has not been a unified study of the three approaches. The purpose of this work was to perform an extensive experimental evaluation of them. For large data sets consisting of originally discrete variables, discretization-based approaches learned the most accurate structures. With smaller sample sizes or data without an underlying discrete mechanism, a method learning directly with continuous data performed best. Also, for some quality metrics, the integrated methods did not provide improvements over simple discretization methods. In terms of time-efficiency, the integrated approaches were the most computationally intensive, while methods from the other categories were the least intensive.