In shotgun proteomics, database search algorithms rely on fragmentation models to predict fragment ions that should be observed for a given peptide sequence. The most widely used strategy (Naïve model) is oversimplified, breaking all peptide bonds with equal probability to produce fragments of all charges below that of the precursor ion. More accurate models are too computationally intensive for on-the-fly use in database search algorithms. We created an ordinal-regression based model called Basophile that reflects the relative importance of basic residues and fragment length in charge retention during CID/HCD fragmentation of charged peptides. The model improves the accuracy of predictions by reducing the number of unnecessary fragments that are routinely predicted for highly charged precursors. When compared with the Naïve model and Protein Prospectors prediction model, Basophile has shown an average of 26% and 28% more identifications in triply-charged precursors on ion trap data. Basophile achieves simplicity and speed by solving the prediction problem with an ordinal regression equation, which can be easily incorporated into any database search software for shotgun proteomic identification.

The utilization of clinical decision support (CDS) is increasing among healthcare facilities that have implemented computerized physician order entry or electronic medical records. Formal prospective evaluation of CDS implementations rarely occurs, and misuse or flaws in system design are often not recognized or corrected. Through retrospective nephrologist adjudication of acute kidney injury (AKI) CDS alerts, we identified patient and knowledgebase factors that contributed to inappropriate, or false positive, alerts. We also estimated the rate of inappropriate provider responses, which occurred in the setting of both true and false positive alerts. We found that few alerts were determined to be inappropriate. Unintended adverse consequences, or inappropriate provider responses resulting from inappropriate alerts, were rare. Retrospective review often occurs too late to make critical corrections or initiate redesign efforts. We developed a real-time, web-based surveillance tool for nephrotoxic and renally cleared medications that integrates provider responses to CDS recommendations with relevant medication ordering, administration, and therapeutic monitoring data. The surveillance view displays all currently admitted, eligible patients and provides brief demographics with triggering order, laboratory, and CDS interactions to facilitate the identification of high-risk patient conditions, such as an imminent adverse drug event (ADE) or potential ADE (pADE). The patient detail view displays a detailed timeline of orders, order administrations, laboratory values, and CDS interactions for an individual patient and allows users to understand provider actions and patient condition changes occurring in conjunction with CDS interactions. We evaluated the surveillance tool with a randomized trial, where intervention patients were monitored on the surveillance tool daily by a clinical pharmacist and control patients received only existing CDS and standard of care. Despite interventions made by the study pharmacist from the surveillance tool, we found no significant change in the timeliness of provider modifications or discontinuations of targeted medications or occurrence of pADEs or ADEs. We concluded that clinical pharmacist surveillance of AKI-related medication alerts did not improve the timeliness or quality of provider responses or patient outcomes.

There is little research on how clinically relevant concepts are transferred from a patient, through a healthcare provider, and then to a resultant clinical note. This study tested whether clinical information flow, defined as the transfer of concepts from patient to note, can be traced. The study's investigators also analyzed the impact of a clinical documentation tool on note content. Healthcare providers, designated as clinical simulation study subjects, generated clinical notes via two documentation tools. The simulation utilized standardized patient scenario descriptions (PSDs). Independent physician reviewers identified clinical concepts present in the PSDs and in resultant clinic notes. Reviewers identified a total of 256 unique clinical concepts across all PSDs. Of these, a total of 122 unique concepts overlapped for the PSDs and resultant notes from both documentation tools. Additionally, the dictation-based and computer-based notes shared 103 distinct concepts not found in the PSDs. This study's findings suggest that both computer-based and dictation-based tools are subject to clinical concept loss. Templates may have eased documentation, partly explaining the greater concept count for computer-based notes. This study found that tracing information flow in a clinical simulation encounter is a valid method for evaluating medical documentation tools. Clinical note template availability also likely impacts healthcare provider documentation.

Shotgun liquid chromatography/tandem mass spectrometry (LC-MS/MS) technology provides data sets rich in the type of information required proteomic quantitation; however, these data are not fully exploited by existing tools. We present a statistical model for combining MS precursor intensity data with MS/MS spectral count data and obtaining a single p-value using Fishers Method of combining p-values. Our model is demonstrated using a new tool, IDPQuantify, which generates MS/MS spectral count data and MS persistent peptide isotopic distribution (PPID) intensity data for peptide group-level difference testing. Using the iPRG 2009 ABRF E. coli data set with known differences in protein content between cohorts, we compared the performance of existing candidate statistical tests using either spectral counts or PPIDs alone. We then compared the performance of our combined model with our candidate tests. Spectral count-based tests showed lower sensitivity but higher specificity than PPID-based tests. In comparison, our combined model yielded a slight drop in sensitivity coupled with an enormous improvement in specificity compared to the PPID-based test alone. We also observed that shared peptide groups tended to yield erroneous rejections of the null hypothesis more often than unshared peptide groups.

Functional Magnetic Resonance Imaging (fMRI) has the potential to unlock many of the mysteries of the brain. Although this imaging modality is popular for brain-mapping activities, clinical applications of this technique are relatively rare. For clinical applications, classification models are more useful than the current practice of reporting loci of neural activation associated with particular disorders. Also, since the methods used to account for anatomical variations between subjects are generally imprecise, the conventional voxel-by-voxel analysis limits the types of discoveries that are possible. This work presents a classification-based framework for knowledge discovery from fMRI data. Instead of voxel-centric knowledge discovery, this framework is segment-centric, where functional segments are clumps of voxels that represent a functional unit in the brain. With simulated activation images, it is shown that this segment-based approach can be more successful for knowledge discovery than conventional voxel-based approaches. The spatial coherence principle refers to the homogeneity of behavior of spatially contiguous voxels. Auto-threshold Contrast Enhancing Iterative Clustering (ACEIC) a new algorithm based on the spatial coherence principle is presented here for functional segmentation. With benchmark data, it is shown that the ACEIC method can achieve higher segmentation accuracy than Probabilistic Independent Component Analysis a popular method used for fMRI data analysis. The spatial coherence principle can also be exploited for voxel-centric image-classification problems. Spatially Coherent Voxels (SCV) is a new feature selection method that uses the spatial coherence principle to eliminate features that are unlikely to be useful for classification. For a Substance Use Disorder dataset, it is demonstrated that feature selection with SCV can achieve higher classification accuracies than conventional feature selection methods.

Background: Disease and patient care processes often create characteristic mathematical and temporal patterns in time-stamped clinical events and observations, but existing medical record systems have a limited ability to recognize or visualize these patterns. System Design: This dissertation introduces the process-oriented approach to clinical data analysis and visualization. This approach aims to support specifying, detecting, and visualizing mathematical and temporal patterns in time-stamped patient data for a broad range of clinical tasks. It has two components: a pattern specification and detection strategy called PROTEMPA (Process-oriented Temporal Analysis); and a pattern visualization strategy called TPOD (Temporal Process-oriented Display). Evaluation: A study in the clinical research domain evaluated PROTEMPAs ability to identify and categorize patients based on diagnosis, disease severity, and disease progression by scanning for patterns in clinical laboratory results. A cognitive study in the patient care domain evaluated PROTEMPA and TPODs ability to help physicians review cases and make decisions using case presentation software that displays laboratory results in either a TPOD-based display or a standard laboratory display. Results: PROTEMPA successfully identified laboratory data patterns in both domains. TPOD successfully visualized these patterns in the patient care domain. In the patient care study, subjects obtained more clinical concepts from the TPOD-based display, but TPOD had no effect on decision-making speed or quality. Subjects were split on which laboratory display they preferred, but expressed a desire to gain more familiarity with the TPOD-based display. Subjects reviewed data in the standard laboratory display for a variety of purposes, and interacted with the display in a complex fashion. Conclusions: The process-oriented approach successfully recognized and visualized data patterns for two distinct clinical tasks. In clinical research, this approach may provide significant advantages over existing methods of data retrieval. In patient care, comparative evaluation of novel data displays in context provides insights into physicians preferences, the process of clinical decision-making by physicians, and display usability. TPODs influence on concept acquisition is promising, but further research is needed regarding physicians use of laboratory data for results review in order to determine how a process-oriented display might be deployed most beneficially.

The goal of the research was to develop a Protein-Specific Properties Evaluation (PSPE) framework that would aid in the statistical evaluation of variables for predicting ranges of and prior probability distributions for protein crystallization conditions. Development of such a framework is motivated by the rapid growth and evolution of the Protein Data Bank. Features of the framework that has been developed include (1) it is an instantiation of the scientific method for the framing and testing of hypotheses in an informatics setting, (2) the use of hidden variables, and (3) a negative result is still useful. The hidden variables examined in this study are related to the estimated net charge (Q) of the proteins under consideration. The Q is a function of the amino acid composition, the solution pH, and the assumed pKa values for the titratable amino acid residues. The proteins size clearly has a significant impact on the magnitude of the Q. Therefore, two additional variables were introduced to mitigate this effect, the specific charge (Qbar) and the average surface charge density (sigma). The principal observation is that proteins appear to crystallize at low values of Qbar and sigma. One problem with this observation is that low is a relative term and the frame of reference requires careful examination. The results are sufficiently weak that no prospective predictions appear possible although information of this type could be included with other weak predictors in a Bayesian predictor scheme. Additional work would be required to establish this; however that work is beyond the scope of the dissertation. Although many statistically significant correlations among Q-related quantities were noted, no evidence could be developed to suggest they were anything other than those expected from the additional information introduced with the hidden variables. Thus, the principal conclusions of this PSPE analysis are that (1) Qbar/sigma and other Q-related variables are of limited value as prospective predictors of ranges of values of crystallization conditions. Although this is a negative result, it is still useful in that it allows attention to be directed into more productive avenues.

The implementation of computerized provider order entry (CPOE) across the health care system has been slow in realization. In addition to the inherent financial burden, a significant cause for this delay is the high number of system failures resulting from clinicians resistance. Changes in workflow and communication, time demands, system complexity, and changes to power structures have all been identified as consequences of CPOE systems that can cause resistance among clinicians. Of these, I believe that perceived changes in a persons power in the workplace can be more difficult to overcome than changes in the work routine. Perception of the power or control that clinicians have in the workplace and their attitudes toward CPOE are precursors to behavior, and if these perceptions and attitudes are negative, can result in resistive behavior. Based on psycho-social theories of power, resistance, and organizational information technology (IT) implementation in business, I applied these concepts to healthcare IT implementation. Qualitative studies have looked at power and resistance, but no previous study has measured the degree or direction of power change, or confirmed that a relationship exists between power perceptions and CPOE attitudes. One reason for this is that no instruments existed to obtain this data. I developed the Semantic Differential Power Perception (SDPP) survey as an electronic survey to measure power perception and CPOE attitudes, and established reliability and validity of the instrument in a measurement study. The SDPP was used to collect data from 276 healthcare workers in two different hospitals before and after implementation of CPOE. I identified a significant correlation between power perceptions and attitudes toward CPOE. Examining the direction of change by healthcare position, we found that the power perception values decreased for all positions and that attitudes toward CPOE varied based on use of the system. Understanding the relationship between power perceptions and CPOE attitudes is the first step in determining causative relationships. This understanding will enable system developers to modify implementation processes and training methods to enhance waning power and support positive power changes, therefore minimizing power related resistance.

Event surveillance involves analyzing a region in order to detect patterns that are indicative of some event of interest. An example is the monitoring of information about emergency department visits to detect a disease outbreak. Spatial event surveillance involves analyzing spatial patterns of evidence that are indicative of the event of interest. A special case of spatial event surveillance is spatial cluster detection, which searches for subregions in which the count of an event of interest is higher than expected. Temporal event surveillance involves monitoring for emerging temporal patterns. Spatio-temporal event surveillance involves joint spatial and temporal monitoring. When the events observed are of direct interest, then analyzing counts of those events is generally the preferred approach. However, in event surveillance we often only observe events that are indirectly related to the events of interest. For example, during an influenza outbreak, we may only have information about the chief complaints of patients who visited emergency departments. In this situation, a better surveillance approach may be to model the relationships among the events of interest and those observed. I developed a high-level Bayesian network architecture that represents a class of spatial event surveillance models, which I call BayesNet-S. I also developed an architecture that represents a class of temporal event surveillance models called BayesNet-T. These Bayesian network architectures are combined into a single architecture that represents a class of spatio-temporal models called BayesNet-ST. Using these architectures, it is often possible to construct a temporal, spatial, or spatio-temporal model from an existing Bayesian network event-surveillance model that is non-spatial and non-temporal. My general hypothesis is that when an existing model is extended to incorporate space and time, event surveillance will be improved. PANDA-CDCA (PC) (Cooper et al., 2007) is a non-temporal, non-spatial disease outbreak detection system. I extended PC both spatially and temporally. My specific hypothesis is that each of the spatial and temporal extensions of PC will perform outbreak detection better than does PC, and that the combined use of the spatial and temporal extensions will perform better than either extension alone. The experimental results obtained in this research support this hypothesis.

High-dimensional biomedical 'omic' datasets are accumulating rapidly from studies aimed at early detection and better management of human disease. These datasets pose tremendous challenges for analysis due to their large number of variables that represent measurements of biochemical molecules, such as proteins and mRNA, from bodily fluids or tissues extracted from a rather small cohort of samples. Machine learning methods have been applied to modeling these datasets including rule learning methods, which have been successful in generating models that are easily interpretable by the scientists. Rule learning methods have typically relied on a frequentist measure of certainty within IF-THEN (propositional) rules. In this dissertation, a Bayesian Rule Generation Framework (BRGF) is developed and tested that can produce rules with probabilities, thereby enabling a mathematically rigorous representation of uncertainty in rule models. The BRGF includes a novel Bayesian Discretization method combined with one or more search strategies for building constrained Bayesian Networks from data and converting them into probabilistic rules. Both global and local structures are built using different Bayesian Network generation algorithms and the rule models generated from the network are tested on public and private 'omic' datasets. We show that using a specific type of structure (Bayesian decision graphs) in tandem with a specific type of search method (parallel greedy) allows us to achieve statistically significant higher overall performance over current state of the art rule learning methods. Not only does using the BRGF boost performance on average on 'omic' biomedical data to a statistically significant point, but also provides the ability to incorporate prior information in a mathematically rigorous fashion for modeling purposes.