Lingsong Zhang

Sparse linear discriminant analysis for simultaneous testing for the significance of a gene set/pathway and gene selection

MOTIVATION Pathway and gene set-based approaches for the analysis of gene expression profiling experiments have become increasingly popular for addressing problems associated with individual gene analysis. Since most genes are not differently expressed, existing gene set tests, which consider all the genes within a gene set, are subject to considerable noise and power loss, a concern exacerbated in studies in which the degree of differential expression is moderate for truly differentially expressed genes. For a significantly differentially expressed pathway, it is also of substantial interest to select important genes that drive the differential expression of the pathway. METHODS We develop a unified framework to jointly test the significance of a pathway and to select a subset of genes that drive the significant pathway effect. To achieve dimension reduction and gene selection, we decompose each gene pathway into a single score by using a regularized form of linear discriminant analysis, called sparse linear discriminant analysis (sLDA). Testing for the significance of the pathway effect proceeds via permutation of the sLDA score. The sLDA-based test is compared with competing approaches with simulations and two applications: a study on the effect of metal fume exposure on immune response and a study of gene expression profiles among Type II Diabetes patients. RESULTS Our results show that sLDA-based testing provides a powerful approach to test for the significance of a differentially expressed pathway and gene selection. AVAILABILITY An implementation of the proposed sLDA-based pathway test in the R statistical computing environment is available at http://www.hsph.harvard.edu/~mwu/software/. SUPPLEMENTARY INFORMATION Supplementary data are available at Bioinformatics online.

Plasma Neutrophil Elastase and Elafin Imbalance Is Associated with Acute Respiratory Distress Syndrome (ARDS) Development

Background We conducted an exploratory study of genome-wide gene expression in whole blood and found that the expression of neutrophil elastase inhibitor (PI3, elafin) was down-regulated during the early phase of ARDS. Further analyses of plasma PI3 levels revealed a rapid decrease during early ARDS development. PI3 and secretory leukocyte proteinase inhibitor (SLPI) are important low-molecular-weight proteinase inhibitors produced locally at neutrophil infiltration site in the lung. In this study, we tested the hypothesis that an imbalance between neutrophil elastase (HNE) and its inhibitors in blood is related to the development of ARDS. Methodology/Principal Findings PI3, SLPI, and HNE were measured in plasma samples collected from 148 ARDS patients and 63 critical ill patients at risk for ARDS (controls). Compared with the controls, the ARDS patients had higher HNE, but lower PI3, at the onset of ARDS, resulting in increased HNE/PI3 ratio (mean = 14.5; 95% CI, 10.9–19.4, P<0.0001), whereas plasma SLPI was not associated with the risk of ARDS development. Although the controls had elevated plasma PI3 and HNE, their HNE/PI3 ratio (mean = 6.5; 95% CI, 4.9–8.8) was not significantly different from the healthy individuals (mean = 3.9; 95% CI, 2.7–5.9). Before the onset (7-days period prior to ARDS diagnosis), we only observed significantly elevated HNE, but the HNE-PI3 balance remained normal. With the progress from prior to the onset of ARDS, the plasma level of PI3 declined, whereas HNE was maintained at a higher level, tilting the balance toward more HNE in the circulation as characterized by an increased HNE/PI3 ratio. In contrast, three days after ICU admission, there was a significant drop of HNE/PI3 ratio in the at-risk controls. Conclusions/Significance Plasma profiles of PI3, HNE, and HNE/PI3 may be useful clinical biomarkers in monitoring the development of ARDS.

Singular Value Decomposition and Its Visualization

Singular value decomposition (SVD) is a useful tool in functional data analysis (FDA). Compared to principal component analysis (PCA), SVD is more fundamental, because SVD simultaneously provides the PCAs in both row and column spaces. We compare SVD and PCA from the FDA view point, and extend the usual SVD to variations by considering different centerings. A generalized scree plot is proposed to select an appropriate centering in practice. Several useful matrix views of the SVD components are introduced to explore different features in data, including SVD surface plots, image plots, curve movies, and rotation movies. These methods visualize both column and row information of a two-way matrix simultaneously, relate the matrix to relevant curves, show local variations, and highlight interactions between columns and rows. Several toy examples are designed to compare the different variations of SVD, and real data examples are used to illustrate the usefulness of the visualization methods.

Some considerations of classification for high dimension low-sample size data.

Abstarct We review in this article several classification methods, especially for high-dimensional and low-sample size data. We discuss several desirable properties for classifiers in such settings, including predictability, consistency, generality, stability, robustness and sparsity. Specifically, a good classifier should have a small prediction error (predictability); converge to the Bayes-rule classifier asymptotically (consistency); be stable when adding/removing an observation (generality); be stable for different data sets of the same kind (stochastic stability); be stable when there are a small number of contaminated observations (robustness); and have a small number of variables in the classifier (interpretability or sparsity). Several simulation examples and real applications are used to illustrate the usefulness of the existing popular classifiers and compare their performance.

No-shows to primary care appointments: subsequent acute care utilization among diabetic patients

BackgroundPatients who no-show to primary care appointments interrupt clinicians’ efforts to provide continuity of care. Prior literature reveals no-shows among diabetic patients are common. The purpose of this study is to assess whether no-shows to primary care appointments are associated with increased risk of future emergency department (ED) visits or hospital admissions among diabetics.MethodsA prospective cohort study was conducted using data from 8,787 adult diabetic patients attending outpatient clinics associated with a medical center in Indiana. The outcomes examined were hospital admissions or ED visits in the 6 months (182 days) following the patient’s last scheduled primary care appointment. The Andersen-Gill extension of the Cox proportional hazard model was used to assess risk separately for hospital admissions and ED visits. Adjustment was made for variables associated with no-show status and acute care utilization such as gender, age, race, insurance and co-morbid status. The interaction between utilization of the acute care service in the six months prior to the appointment and no-show was computed for each model.ResultsThe six-month rate of hospital admissions following the last scheduled primary care appointment was 0.22 (s.d. = 0.83) for no-shows and 0.14 (s.d. = 0.63) for those who attended (p < 0.0001). No-show was associated with greater risk for hospitalization only among diabetics with a hospital admission in the prior six months. Among diabetic patients with a prior hospital admission, those who no-showed were at 60% greater risk for subsequent hospital admission (HR = 1.60, CI = 1.17–2.18) than those who attended their appointment. The six-month rate of ED visits following the last scheduled primary care appointment was 0.56 (s.d. = 1.48) for no-shows and 0.38 (s.d. = 1.05) for those who attended (p < 0.0001); after adjustment for covariates, no-show status was not significantly related to subsequent ED utilization.ConclusionsNo-show to a primary care appointment is associated with increased risk for hospital admission among diabetics recently hospitalized.

Robust regularized singular value decomposition with application to mortality data

We develop a robust regularized singular value decomposition (RobRSVD) method for analyzing two-way functional data. The research is motivated by the application of modeling human mortality as a smooth two-way function of age group and year. The RobRSVD is formulated as a penalized loss minimization problem where a robust loss function is used to measure the reconstruction error of a low-rank matrix approximation of the data, and an appropriately defined two-way roughness penalty function is used to ensure smoothness along each of the two functional domains. By viewing the minimization problem as two conditional regularized robust regressions, we develop a fast iterative reweighted least squares algorithm to implement the method. Our implementation naturally incorporates missing values. Furthermore, our formulation allows rigorous derivation of leave-one-row/column-out cross-validation and generalized cross-validation criteria, which enable computationally efficient data-driven penalty parameter selection. The advantages of the new robust method over nonrobust ones are shown via extensive simulation studies and the mortality rate application.

No-Show Modeling for Adult Ambulatory Clinics

Patient no-show is a pervasive problem in outpatient clinics. This chapter provides a literature review and discussion on how to develop statistical no-show models. The literature review is a structured and representative selection of research studies from a variety of medical areas. The literature is grouped into four classes. The first class covers self-reported reasons for no-show. The most common self-reported reasons are forgetting, conflicts, transportation, scheduling system problems, and physical or mental illness. The second class discusses the effect of no-show interventions such as appointment reminders, patient education, and changes in scheduling systems on no-show behavior. The third class develops statistical models of no-show behavior in a variety of settings. Several patient, provider, and clinic characteristics are considered in developing these models. The last class of literature considers the impact of no-shows on health outcomes, which illustrates the importance of no-show modeling. The second part of the chapter explains how statistical no-show models can be developed. The data requirements, determination of significant factors, development of logistic regression models, and model validation are explained in detail. An example no-show model is provided to illustrate the modeling and validation process. The chapter concludes with summarizing thoughts and a discussion of future research opportunities.

The impact of interventions on appointment and clinical outcomes for individuals with diabetes: a systematic review.

BackgroundSuccessful diabetes disease management involves routine medical care with individualized patient goals, self-management education and on-going support to reduce complications. Without interventions that facilitate patient scheduling, improve attendance to provider appointments and provide patient information to provider and care team, preventive services cannot begin. This review examines interventions based upon three focus areas: 1) scheduling the patient with their provider; 2) getting the patient to their appointment, and; 3) having patient information integral to their diabetes care available to the provider. This study identifies interventions that improve appointment management and preparation as well as patient clinical and behavioral outcomes.MethodsA systematic review of the literature was performed using MEDLINE, CINAHL and the Cochrane library. Only articles in English and peer-reviewed articles were chosen. A total of 77 articles were identified that matched the three focus areas of the literature review: 1) on the schedule, 2) to the visit, and 3) patient information. These focus areas were utilized to analyze the literature to determine intervention trends and identify those with improved diabetes clinical and behavioral outcomes.ResultsThe articles included in this review were published between 1987 and 2013, with 46 of them published after 2006. Forty-two studies considered only Type 2 diabetes, 4 studies considered only Type 1 diabetes, 15 studies considered both Type 1 and Type 2 diabetes, and 16 studies did not mention the diabetes type. Thirty-five of the 77 studies in the review were randomized controlled studies. Interventions that facilitated scheduling patients involved phone reminders, letter reminders, scheduling when necessary while monitoring patients, and open access scheduling. Interventions used to improve attendance were letter reminders, phone reminders, short message service (SMS) reminders, and financial incentives. Interventions that enabled routine exchange of patient information included web-based programs, phone calls, SMS, mail reminders, decision support systems linked to evidence-based treatment guidelines, registries integrated with electronic medical records, and patient health records.ConclusionsThe literature review showed that simple phone and letter reminders for scheduling or prompting of the date and time of an appointment to more complex web-based multidisciplinary programs with patient self-management can have a positive impact on clinical and behavioral outcomes for diabetes patients. Multifaceted interventions aimed at appointment management and preparation during various phases of the medical outpatient care process improves diabetes disease management.

Multi-Resolution Anomaly Detection for the internet

In the context of Internet traffic anomaly detection, we will show that some outliers in a time series can be difficult to detect at one scale while they are easy to find at another scale. In this paper, we develop an outlier detection method for a time series with long range dependence, and conclude that testing outliers at multiple time scales helps to reveal them. We present a multi-resolution anomaly detection (MRAD) procedure for detecting network anomalies. We show that the MRAD method is useful, especially when outliers appear as a slight local mean level shift with a rather long duration, e.g., as generated by a port scan. A novel MRAD outlier map is proposed to visualize the location of the outliers, and also to suggest the significance probabilities (p values) for them.

The Effect of Interactive Web-Based Monitoring on Breastfeeding Exclusivity, Intensity, and Duration in Healthy, Term Infants After Hospital Discharge

OBJECTIVE To determine whether a Web-based interactive breastfeeding monitoring system increased breastfeeding duration, exclusivity, and intensity as primary outcomes and decreased symptoms of postpartum depression as a secondary outcome. DESIGN Two-arm, randomized controlled trial. SETTING Three hospitals in the Midwestern United States. PARTICIPANTS One hundred forty one (141) mother-newborn dyads were recruited before discharge. METHODS Postpartum women were randomly assigned to the control or intervention groups. Women in the control group (n = 57) followed the standard hospital protocol, whereas women in the intervention group (n = 49) were given access to an online interactive breastfeeding monitoring system and were prompted to record breastfeeding and infant output data for 30 days. A follow-up online survey was sent to both groups at 1, 2, and 3 months to assess breastfeeding outcomes and postpartum depression. RESULTS For mothers and infants, there were no significant differences in demographics between groups. No significant differences in breastfeeding outcomes were found between groups at discharge (p = .707). A significant difference in breastfeeding outcomes was found between groups at 1, 2, and 3 months (p = .027, p < .001, and p = .002, respectively). Members of the intervention group had greater exclusive breastfeeding rates at 1, 2, and 3 months. By the end of the third month, 84% of the intervention group was breastfeeding compared with 66% of the control group. Postpartum depression symptom scores decreased for both groups at 1, 2, and 3 months (control group: 4.9 ± 3.9, 4.3 ± 4.9, and 3.2 ± 3.9, respectively; intervention group: 4.7 ± 4.5, 3.0 ± 3.4, and 2.8 ± 3.6, respectively). However, there was no significant difference between groups at 1, 2, and 3 months (p = .389, .170, and .920, respectively) for depression. CONCLUSION The Web-based interactive breastfeeding monitoring system may be a promising intervention to improve breastfeeding duration, exclusivity, and intensity.