Sudden Infant Death (SIDS) has been associated with medium-chain acyl-CoA dehydrogenase (MCAD) deficiency, an inborn error of fatty acid oxidation. Blood and tissue samples from a large cohort of SIDS victims were analyzed for the presence of dodecanoic acid (C₁₂) by gas chromatography. A subgroup of these cases had a significantly higher blood concentration than age-matched controls, suggesting MCAD deficiency. An animal study using Sprague-Dawley rats was done to mimic the effects of MCAD deficiency. Significantly increased blood concentrations of dodecanoic acid were observed. Decreased values in heart and liver were puzzling findings. The data indicate that dodecanoic acid is a blood marker for MCAD deficiency.

In the realm of biomedical technology, both accuracy and consistency are crucial to the development and deployment of these tools. While accuracy is easy to measure, consistency metrics are not so simple to measure, especially in the scope of biomedicine where prediction consistency can be difficult to achieve. Typically, biomedical datasets contain a significantly larger amount of features compared to the amount of samples, which goes against ordinary data mining practices. As a result, predictive models may fail to find valid pathways for prediction during training on such datasets. This concept is known as underspecification. Underspecification has been more accepted as a concept in recent years, with a handful of recent works exploring underspecification in different applications and a handful of past works experiencing underspecification prior to its declaration. However, underspecification is still under-addressed, to the point where some academics might even claim that it is not a significant problem. With this in mind, this thesis aims to identify and minimize underspecification of deep learning cancer subtype predictors. To address these goals, this work details the development of Predicting Underspecification Monitoring Pipeline (PUMP), a software tool to provide methodology for data analysis, stress testing, and model evaluation. In this context, the hope is that PUMP can be applied to deep learning training such that any user can ensure that their models are able to generalize to new data as best as possible.

No description available.

Maintaining balance in the human immune system is critical for the body’s ability to discriminate between foreign and self-antigens. This balance is achieved, in part, by a subpopulation of T cells known as induced regulatory T cells (iTregs). Dysregulation of this population may contribute to the onset and progression of cancer, chronic inflammation and autoimmune diseases. Therefore, manipulation of iTreg development holds promising therapeutic potential; however, studying this vital population has proven difficult due to low numbers, heterogeneous cell populations, substantial phenotypic differences between mouse and human cells, and the high plasticity seen in iTregs. These current limitations have prevented a full understanding of the molecular signaling events that govern their development and function. Our lab has established a novel cell culture system that mimics in vivo human iTreg development. This system allows for the discrimination and comparison of naïve, memory and iTreg T cell populations simultaneously within a single donor. These iTregs exhibit high levels of CD25, FoxP3, CTLA4, GITR, low levels of CD127 and display strong suppressor activity. Using this innovative system, we have demonstrated a rewiring of T cell receptor (TCR) signaling in iTregs compared to conventional T cells. We found that the voltage gated K+ ion channel-Kv1.3 is not active in response to TCR engagement in iTregs, even though Ca2+ influx remains intact. Kv1.3 and the linked Src-family kinase Lck were redistributed to the highly active IL2-Receptor (IL2-R) complex. Additionally, we have shown that there is increased AKT protein expression in iTregs versus conventional T cell populations that does not correlate with the TCR-induced increase in its active (phosphorylated) form. This blockage appears to be due to an imbalance of kinase to phosphatase activity in iTregs with a specific TCR-induced inhibition of mTOR activity. We have also demonstrated that AKT accumulation in iTregs leads to its physical association with SMAD3, suggesting a novel, non-enzymatic function of AKT through transcription factor inhibition. This study sheds light on the reciprocal cross talk between the IL-2R and TCR signaling pathways and uncovers the mechanism of AKT blockade in primary human iTregs, thus opening novel avenues for therapeutic manipulation

Preeclampsia (PE), characterized by hypertension and proteinuria after 20 weeks of gestation, affects 5-8% of pregnancies worldwide. Although preeclampsia is a significant cause of maternal and perinatal mortality and morbidity, its etiology remains to be elucidated. Racial differences have been observed for preeclampsia, with U.S. Blacks having higher rates and more severe disease, compared to U.S. Whites and Hispanics. One potential source of racial differences in preeclampsia is genetic variation between populations. Genetic susceptibility to preeclampsia is well established, but the specific contributions of maternal vs. fetal genes, and how these vary among racial groups is poorly understood. This dissertation addressed racial differences in the genetics of preeclampsia in Chileans, U.S. Blacks, and U.S. Whites through candidate gene studies and variance components modeling. First, we determined whether three genes, which are relevant to the pathophysiology of preeclampsia, Catechol-O-methyltransferase (COMT), Methylenetetrahydrofolate reductase (MTHFR), and Endoplasmic reticulum aminopeptidase 2 (ERAP2), were associated with the risk for preeclampsia in Chilean and U.S. Black mothers and fetuses. We found that the maternal COMT and an interaction between the fetal COMT and MTHFR were associated with the risk for preeclampsia in Chileans. We also found that the fetal ERAP2 was associated with the risk for preeclampsia in U.S. Blacks. We next used structural equation modeling of a unique Children of Twins (COT), supplemented with full and half-siblings, study design to investigate the fetal genetic, maternal genetic, shared environmental, and unique environmental contributions to preeclampsia in U.S. Whites and Blacks. Through this modeling we uncovered a unique source of racial differences in preeclampsia. We found that U.S. Whites and Blacks showed a similar prevalence of preeclampsia in first births, but across the next three births, the prevalence in Whites declined to a greater degree than in Blacks. In conclusion we have identified specific maternal and fetal genes that contribute to the risk for preeclampsia. Furthermore, we have identified sources of racial differences in preeclampsia, which include differences in associations between COMT, MTHFR, and ERAP2 and the risk for preeclampsia among populations and differences in the prevalence of preeclampsia across subsequent births between U.S. Whites and U.S. Blacks.

Obesity is a serious public health crisis and recent estimates of its incidence are the highest in United States history, with 35% and 17% of American adults and children affected, respectively. The clinical definition of adult obesity is operationalized as a body mass index (BMI) greater than 30 kg/m2. Although the prevalence of common obesity has increased dramatically over the past 30 years–largely thought to be due to changes in the environment, such as high calorie diets and sedentary lifestyles—twin and family studies have shown consistently that relative body weight is under considerable genetic influence in both children and adults, with heritability estimates ranging from 40% to 90%. Elucidating the genetic and environmental liability to relative body weight is an important public health endeavor. To further our understanding of the genetics of BMI and common complex obesity, several studies are described that integrate clinical, twin, and genome-wide association (GWAS) methodology in the context of genetic risk scores, clinical risk prediction, development across adolescence into adulthood, and comorbidity with depression symptoms and smoking behavior. First, in two cross-sectional genetic association studies, the utility of genetic risk sum scores (GRSS) were assessed, which summarize the total number of risk alleles, as an alternative form of replication and for potential clinical utility for obesity risk prediction. Next, since there has been only limited research on when during development BMI-associated variants begin influencing BMI, a longitudinal twin study was utilized to assess the effects of adult-validated BMI-SNPs across adolescence into adulthood. In addition, obesity is comorbid with numerous medical conditions including cardiovascular disease, insulin-resistance and some forms of cancer, as well as, various psychiatric disorders including eating disorders, mood disorders, and substance use. The next series of studies aimed to understand phenotypic and genetic associations between BMI/obesity and binge eating disorder (BED), depression symptoms and smoking behavior. Using a clinical sample of overweight and obese women with and without BED, the relationship of BED, food intake and internalizing symptoms of depression and anxiety was examined. Next, twin study methodology was used to investigate if shared genetic and/or environmental liability was responsible for phenotypic associations found between BMI, depression symptoms, and impulsivity. Finally, a genetic association study aimed at investigating whether genetic variants were associated with multiple behaviors, body composition and smoking behavior, or were trait-specific is presented. By utilizing several samples and methodologies and by pursuing methods development, a comprehensive approach is presented that is hoped to represent a more powerful evidence-based strategy to understanding the genetic and environmental determinants of BMI and common complex obesity, along with associated depression symptoms and smoking behavior.

Schizophrenia is a complex disorder affecting 1% of the population and is highly heritable, but the majority of contributing genetic factors has remained elusive. Current risk estimates for clinical practice are primarily determined by family history and associated empirical risk. Copy number variants (CNVs) may hold the key to explaining the missing heritability in schizophrenia research; schizophrenia risk estimates as high as 30% have been found for the most-studied CNV associated with schizophrenia, 22q11. Currently, there are methods to identify CNVs though previously collected data from SNP microarrays that would facilitate these types of studies. To determine if algorithms that call CNVs from microarray data are robust four genomic regions with putative CNVs called by the Wellcome Trust Consortium using Birdseye in Birdsuite with Affymetrix 6.0 array raw SNP intensities, primarily affecting genes CHD1L, COX5B, PAK7, ZFYVE20, were validated using Taqman real-time qPCR assays in 29 samples by research groups at VCU and Dublin. CNVs called from the algorithm were 100% validated at VCU though there were false negatives from the algorithm that were validated. Two samples at loci with putative duplications were not called by the Dublin group, which may be because of differing sensitivities of the Taqman assays to be able to detect a 50% difference in copy number between duplications and diploid controls, or because of another technical or analytical difference between the two sites. Deletion frequency of one common CNV found in the gene ERBB4, was assessed by qPCR in both Irish singleton (ICCSS) and Irish family (IHDSF) samples and compared with Irish control (Trinity Biobank) and North American control populations. The ERBB4 deletion frequency was not significantly different when comparing the Irish controls to the Irish singleton or the Irish family samples though the family samples were different when compared against the North American control population, which suggests population stratification, rather than a true association between ERBB4 and increased schizophrenia risk. Current clinical practice has been improved by knowledge and evaluation of CNV-related disorders that include risk for psychosis and additional phenotypes. Genotyping of individuals with known psychosis has led to improved patient care for non-psychosis-related phenotypes associated with CNVs. Individuals with suspected genomic disorders that are found to have CNVs can be counseled on potential psychosis risk and potential risk to their offspring. Recurrent CNVs may hold promise in future clinical practice in order to individualize risk estimates in the general patient population, and increase the number of individuals able to receive anticipatory treatment to minimize disease severity.

Smith-Magenis syndrome (SMS) is a complex disorder caused by haploinsufficiency of RAI1 and characterized by sleep disturbances, behavioral abnormalities, mental retardation, and obesity in teens and adults. Rai1+/- mice are obese after 20 weeks. Dup(17)(p11.2) syndrome is a complex disorder associated with overexpression of RAI1. A transgenic mouse model of dup(17)(p11.2) syndrome overexpresses Rai1 and results in a mouse that is growth delayed. In order to characterize the obese phenotypes of mouse models of SMS and the role of RAI1 in obesity, daily food intake and serum levels of insulin, glucose, PPY, and leptin were measured; adiposity was studied by characterizing fat deposition; and gene expression was studied in the hypothalamus. These studies show that Rai1+/- mice are hyperphagic, consume more during the inactive light phase, and have altered satiety genes in the hypothalamus. Adiposity studies have shown WT females have a higher body fat content and visceral fat proportion than males, but Rai1-Tg and Rai1+/- females have similar fat deposition patterns as WT males. Hypothalamic gene expression studies show that many genes and pathways are affected by Rai1 and Rai1 dosage, including many genes associated with obesity and satiety.

Increasingly the role of epigenetic machinery as a bridge between underlying DNA sequence and cellular phenotype is being discovered. The establishment of a myriad of unique cellular types sharing identical gene sequences in a multicellular organism gives a broad sense for the inherent role of epigenetic influence on cell differentiation. Importantly, the epigenetic mechanisms involved in establishing cell identity unsurprisingly contribute to diseased states, including cancer. Recent research continues to elucidate contributory roles of epigenetic mechanisms, such as DNA methylation, histone modification, and microRNA regulation, in human cancers. Additionally, chromatin remodelers, such as the Nucleosome Remodeling Factor (NURF), have been identified as important regulators for normal cell biology. While much has been done to identify and characterize the role of NURF chromatin remodeling complex as a key regulator of development in a number of model organisms, little has been published on the implications of NURF in diseases such as cancer. Our preliminary data shows dysregulation of E-cadherins, N-cadherins, and MHC-I genes in Bptf (an essential subunit of NURF) knocked down murine breast cancer cell lines. These proteins have well documented roles in the development and metastatic progression of cancers. To study the effect of Bptf knockdown on the development and progression of cancer we injected Bptf knocked down mouse breast cancer cell lines, 4T1, 66cl4, and 67NR, into syngenic BALB/c mice. Our findings reveal decreased tumor growth in 66cl4 and 67NR as measured by tumor weight at 3-4 weeks post injection. Tumor growth did not appear to be significantly affected in 4T1 challenged mice. However, mice inoculated with Bptf knockdown 4T1 cell lines have decreased metastasis to lungs as compared to control while metastasis of 66cl4 tumors to the lungs appear unaffected. To assess the role of the immune system in decreasing tumor growth in BALB/c mice, we injected 66cl4 tumors into NOD-SCID-Gamma (NSG) immune deficient mice. The tumors from these mice show no difference in tumor growth between Bptf knockdown and control tumors, implicating a role for the immune system regulating the decreased tumor weight in BALB/c mice. To delineate which immune cell effector may impede breast cancer carcinogenesis, we performed an in vitro natural killer (NK) cell cytotoxicity assay against 66cl4 tumors and found greater susceptibility to NK killing in Bptf knockdown tumors.

Alcohol dependence (AD) is a serious and common public health problem that contributes to great societal, medical, and legal costs. It has taken work from multiple disciplines, including developmental psychology, genetic epidemiology, and molecular genetics, to achieve our current understanding of environmental and genetic risk factors for AD as well as its variable developmental trajectories. Nevertheless, there is still much to be learned in order to improve treatment outcomes. One approach to augmenting our understanding of this disorder is through genetically informative study designs that either examine risk in aggregate or assess specific susceptibility variants. In this dissertation, we utilize both study designs and provide support for the idea that they are both important and useful approaches to continue to pursue.