Comparison of multiple comparison methods for identifying differential gene expression in simulated and real papillary thyroid cancer microarray data

The difficulty of detecting differential gene expression in microarray data has existed for many years. Several correction procedures try to avoid the family-wise error rate in multiple comparison process, including the Bonferroni and Sidak single-step p-value adjustments, Holm's step-down correction method, and Benjamini and Hochberg's false discovery rate (FDR) correction procedure. Each multiple comparison technique has its advantages and weaknesses. We studied each multiple comparison method through numerical studies (simulations) and applied the methods to the real exploratory DNA microarray data, which detect of molecular signatures in papillary thyroid cancer (PTC) patients. According to our results of simulation studies, Benjamini and Hochberg step-up FDR controlling procedure is the best process among these multiple comparison methods and we discovered 1277 potential biomarkers among 54675 probe sets after applying the Benjamini and Hochberg's method to PTC microarray data.^

Statistical Methods for Differential Expressions of Genes Detected in Multiple-Condition Experiment of Microarray

Most studies of differential gene-expressions have been conducted between two given conditions. The two-condition experimental (TCE) approach is simple in that all genes detected display a common differential expression pattern responsive to a common two-condition difference. Therefore, the genes that are differentially expressed under the other conditions other than the given two conditions are undetectable with the TCE approach. In order to address the problem, we propose a new approach called multiple-condition experiment (MCE) without replication and develop corresponding statistical methods including inference of pairs of conditions for genes, new t-statistics, and a generalized multiple-testing method for any multiple-testing procedure via a control parameter C. We applied these statistical methods to analyze our real MCE data from breast cancer cell lines and found that 85 percent of gene-expression variations were caused by genotypic effects and genotype-ANAX1 overexpression interactions, which agrees well with our expected results. We also applied our methods to the adenoma dataset of Notterman et al. and identified 93 differentially expressed genes that could not be found in TCE. The MCE approach is a conceptual breakthrough in many aspects: (a) many conditions of interests can be conducted simultaneously; (b) study of association between differential expressions of genes and conditions becomes easy; (c) it can provide more precise information for molecular classification and diagnosis of tumors; (d) it can save lot of experimental resources and time for investigators.^

Effect of insurance status on stage of breast cancer diagnosis

Objective: The purpose of this study is to compare the stages of breast cancer presented between the insured and uninsured patients diagnosed at The Rose, an active non-profit breast healthcare organization to determine if uninsured patients present with more advanced stage breast cancer as compared to their insured counterparts. ^ Study Design: Retrospective cross-sectional study. ^ Methods: The study included 1,265 patients who received breast healthcare services and were diagnosed with breast cancer at The Rose between FY 2007 and FY 2012. 738 of the patients in the study were presumably uninsured since their breast healthcare services were sponsored through various funding sources and they were navigated into treatment through The Rose patient navigation program. We compared breast cancer stages for women who had insurance with those who did not have insurance. The effects of age and race/ethnicity along with the insurance status on the stage of reast cancer diagnosis were also analyzed. We calculated the odds ratio using the contingency tables; and estimated odds ratios (ORs) and 95% confidence intervals (CIs) using ordinal logistic regression by applying multiple imputation method for missing tumor stage data. ^ Results: The ordered logistic regression analysis with ordered tumor stage as dependent variable and uninsured as independent variable gave us an odds ratio of 1.73 (OR=1.73; p-value<0.05; 95% CI: 1.36 - 2.12). ^ Conclusions: Insurance status is a strong predictor of stage of breast cancer diagnosed among women seen at The Rose. Uninsured women seen at The Rose are almost twice as likely to present at a advanced stage of breast cancer as opposed to their insured counterparts.^

Automatic quantitation of multiple sclerosis lesions on MR images

A two-pronged approach for the automatic quantitation of multiple sclerosis (MS) lesions on magnetic resonance (MR) images has been developed. This method includes the design and use of a pulse sequence for improved lesion-to-tissue contrast (LTC) and seeks to identify and minimize the sources of false lesion classifications in segmented images. The new pulse sequence, referred to as AFFIRMATIVE (Attenuation of Fluid by Fast Inversion Recovery with MAgnetization Transfer Imaging with Variable Echoes), improves the LTC, relative to spin-echo images, by combining Fluid-Attenuated Inversion Recovery (FLAIR) and Magnetization Transfer Contrast (MTC). In addition to acquiring fast FLAIR/MTC images, the AFFIRMATIVE sequence simultaneously acquires fast spin-echo (FSE) images for spatial registration of images, which is necessary for accurate lesion quantitation. Flow has been found to be a primary source of false lesion classifications. Therefore, an imaging protocol and reconstruction methods are developed to generate "flow images" which depict both coherent (vascular) and incoherent (CSF) flow. An automatic technique is designed for the removal of extra-meningeal tissues, since these are known to be sources of false lesion classifications. A retrospective, three-dimensional (3D) registration algorithm is implemented to correct for patient movement which may have occurred between AFFIRMATIVE and flow imaging scans. Following application of these pre-processing steps, images are segmented into white matter, gray matter, cerebrospinal fluid, and MS lesions based on AFFIRMATIVE and flow images using an automatic algorithm. All algorithms are seamlessly integrated into a single MR image analysis software package. Lesion quantitation has been performed on images from 15 patient volunteers. The total processing time is less than two hours per patient on a SPARCstation 20. The automated nature of this approach should provide an objective means of monitoring the progression, stabilization, and/or regression of MS lesions in large-scale, multi-center clinical trials. ^

A novel treatment planning methodology for high dose 166Ho-DOTMP therapy in patients with multiple myeloma

Bone marrow ablation, i.e., the complete sterilization of the active bone marrow, followed by bone marrow transplantation (BMT) is a comment treatment of hematological malignancies. The use of targeted bone-seeking radiopharmaceuticals to selectively deliver radiation to the adjacent bone marrow cavities while sparing normal tissues is a promising technique. Current radiopharmaceutical treatment planning methods do not properly compensate for the patient-specific variable distribution of radioactive material within the skeleton. To improve the current method of internal dosimetry, novel methods for measuring the radiopharmaceutical distribution within the skeleton were developed. 99mTc-MDP was proven as an adequate surrogate for measuring 166Ho-DOTMP skeletal uptake and biodistribution, allowing these measures to be obtained faster, safer, and with higher spatial resolution. This translates directly into better measurements of the radiation dose distribution within the bone marrow. The resulting bone marrow dose-volume histograms allow prediction of the patient disease response where conventional organ scale dosimetry failed. They indicate that complete remission is only achieved when greater than 90% of the bone marrow receives at least 30 Gy. ^ Comprehensive treatment planning requires combining target and non-target organ dosimetry. Organs in the urinary tract were of special concern. The kidney dose is primarily dependent upon the mean transit time of 166 Ho-DOTMP through the kidney. Deconvolution analysis of renograms predicted a mean transit time of 2.6 minutes for 166Ho-DOTMP. The radiation dose to the urinary bladder wall is dependent upon numerous factors including patient hydration and void schedule. For beta-emitting isotopes such as 166Ho, reduction of the bladder wall dose is best accomplished through good patient hydration and ensuring a partially full bladder at the time of injection. Encouraging the patient to void frequently, or catheterizing the patient without irrigation, will not significantly reduce the bladder wall dose. ^ The results from this work will produce the most advanced treatment planning methodology for bone marrow ablation therapy using radioisotopes currently available. Treatments can be tailored specifically for each patient, including the addition of concomitant total body irradiation for patients with unfavorable dose distributions, to deliver a desired patient disease response, while minimizing the dose or toxicity to non-target organs. ^

Multiple gradient echo propeller (MGREP) MRI: Technical development and potential applications

The PROPELLER (Periodically Rotated Overlapping Parallel Lines with Enhanced Reconstruction) magnetic resonance imaging (MRI) technique has inherent advantages over other fast imaging methods, including robust motion correction, reduced image distortion, and resistance to off-resonance effects. These features make PROPELLER highly desirable for T2*-sensitive imaging, high-resolution diffusion imaging, and many other applications. However, PROPELLER has been predominantly implemented as a fast spin-echo (FSE) technique, which is insensitive to T2* contrast, and requires time-inefficient signal averaging to achieve adequate signal-to-noise ratio (SNR) for many applications. These issues presently constrain the potential clinical utility of FSE-based PROPELLER. ^ In this research, our aim was to extend and enhance the potential applications of PROPELLER MRI by developing a novel multiple gradient echo PROPELLER (MGREP) technique that can overcome the aforementioned limitations. The MGREP pulse sequence was designed to acquire multiple gradient-echo images simultaneously, without any increase in total scan time or RF energy deposition relative to FSE-based PROPELLER. A new parameter was also introduced for direct user-control over gradient echo spacing, to allow variable sensitivity to T2* contrast. In parallel to pulse sequence development, an improved algorithm for motion correction was also developed and evaluated against the established method through extensive simulations. The potential advantages of MGREP over FSE-based PROPELLER were illustrated via three specific applications: (1) quantitative T2* measurement, (2) time-efficient signal averaging, and (3) high-resolution diffusion imaging. Relative to the FSE-PROPELLER method, the MGREP sequence was found to yield quantitative T2* values, increase SNR by ∼40% without any increase in acquisition time or RF energy deposition, and noticeably improve image quality in high-resolution diffusion maps. In addition, the new motion algorithm was found to improve the performance considerably in motion-artifact reduction. ^ Overall, this work demonstrated a number of enhancements and extensions to existing PROPELLER techniques. The new technical capabilities of PROPELLER imaging, developed in this thesis research, are expected to serve as the foundation for further expanding the scope of PROPELLER applications. ^

The information bottleneck method for genome-wide association studies

In population studies, most current methods focus on identifying one outcome-related SNP at a time by testing for differences of genotype frequencies between disease and healthy groups or among different population groups. However, testing a great number of SNPs simultaneously has a problem of multiple testing and will give false-positive results. Although, this problem can be effectively dealt with through several approaches such as Bonferroni correction, permutation testing and false discovery rates, patterns of the joint effects by several genes, each with weak effect, might not be able to be determined. With the availability of high-throughput genotyping technology, searching for multiple scattered SNPs over the whole genome and modeling their joint effect on the target variable has become possible. Exhaustive search of all SNP subsets is computationally infeasible for millions of SNPs in a genome-wide study. Several effective feature selection methods combined with classification functions have been proposed to search for an optimal SNP subset among big data sets where the number of feature SNPs far exceeds the number of observations. ^ In this study, we take two steps to achieve the goal. First we selected 1000 SNPs through an effective filter method and then we performed a feature selection wrapped around a classifier to identify an optimal SNP subset for predicting disease. And also we developed a novel classification method-sequential information bottleneck method wrapped inside different search algorithms to identify an optimal subset of SNPs for classifying the outcome variable. This new method was compared with the classical linear discriminant analysis in terms of classification performance. Finally, we performed chi-square test to look at the relationship between each SNP and disease from another point of view. ^ In general, our results show that filtering features using harmononic mean of sensitivity and specificity(HMSS) through linear discriminant analysis (LDA) is better than using LDA training accuracy or mutual information in our study. Our results also demonstrate that exhaustive search of a small subset with one SNP, two SNPs or 3 SNP subset based on best 100 composite 2-SNPs can find an optimal subset and further inclusion of more SNPs through heuristic algorithm doesn't always increase the performance of SNP subsets. Although sequential forward floating selection can be applied to prevent from the nesting effect of forward selection, it does not always out-perform the latter due to overfitting from observing more complex subset states. ^ Our results also indicate that HMSS as a criterion to evaluate the classification ability of a function can be used in imbalanced data without modifying the original dataset as against classification accuracy. Our four studies suggest that Sequential Information Bottleneck(sIB), a new unsupervised technique, can be adopted to predict the outcome and its ability to detect the target status is superior to the traditional LDA in the study. ^ From our results we can see that the best test probability-HMSS for predicting CVD, stroke,CAD and psoriasis through sIB is 0.59406, 0.641815, 0.645315 and 0.678658, respectively. In terms of group prediction accuracy, the highest test accuracy of sIB for diagnosing a normal status among controls can reach 0.708999, 0.863216, 0.639918 and 0.850275 respectively in the four studies if the test accuracy among cases is required to be not less than 0.4. On the other hand, the highest test accuracy of sIB for diagnosing a disease among cases can reach 0.748644, 0.789916, 0.705701 and 0.749436 respectively in the four studies if the test accuracy among controls is required to be at least 0.4. ^ A further genome-wide association study through Chi square test shows that there are no significant SNPs detected at the cut-off level 9.09451E-08 in the Framingham heart study of CVD. Study results in WTCCC can only detect two significant SNPs that are associated with CAD. In the genome-wide study of psoriasis most of top 20 SNP markers with impressive classification accuracy are also significantly associated with the disease through chi-square test at the cut-off value 1.11E-07. ^ Although our classification methods can achieve high accuracy in the study, complete descriptions of those classification results(95% confidence interval or statistical test of differences) require more cost-effective methods or efficient computing system, both of which can't be accomplished currently in our genome-wide study. We should also note that the purpose of this study is to identify subsets of SNPs with high prediction ability and those SNPs with good discriminant power are not necessary to be causal markers for the disease.^

A novel method for evaluating an interaction effect of correlated continuous covariates in a linear model

Interaction effect is an important scientific interest for many areas of research. Common approach for investigating the interaction effect of two continuous covariates on a response variable is through a cross-product term in multiple linear regression. In epidemiological studies, the two-way analysis of variance (ANOVA) type of method has also been utilized to examine the interaction effect by replacing the continuous covariates with their discretized levels. However, the implications of model assumptions of either approach have not been examined and the statistical validation has only focused on the general method, not specifically for the interaction effect.^ In this dissertation, we investigated the validity of both approaches based on the mathematical assumptions for non-skewed data. We showed that linear regression may not be an appropriate model when the interaction effect exists because it implies a highly skewed distribution for the response variable. We also showed that the normality and constant variance assumptions required by ANOVA are not satisfied in the model where the continuous covariates are replaced with their discretized levels. Therefore, naïve application of ANOVA method may lead to an incorrect conclusion. ^ Given the problems identified above, we proposed a novel method modifying from the traditional ANOVA approach to rigorously evaluate the interaction effect. The analytical expression of the interaction effect was derived based on the conditional distribution of the response variable given the discretized continuous covariates. A testing procedure that combines the p-values from each level of the discretized covariates was developed to test the overall significance of the interaction effect. According to the simulation study, the proposed method is more powerful then the least squares regression and the ANOVA method in detecting the interaction effect when data comes from a trivariate normal distribution. The proposed method was applied to a dataset from the National Institute of Neurological Disorders and Stroke (NINDS) tissue plasminogen activator (t-PA) stroke trial, and baseline age-by-weight interaction effect was found significant in predicting the change from baseline in NIHSS at Month-3 among patients received t-PA therapy.^