Search

Sequence classification is an important problem in many real-world applications. Unlike other machine learning data, there are no "explicit" features or signals in sequence data that can help traditional machine learning algorithms learn and predict from the data. Sequence data exhibits inter-relationships in the elements that are important in understanding and predicting future sequences. However, finding these relationships is proven to be an NPhard problem. When we use naive enumerations of combinations of elements or "brute force" iterative approaches for defining these features they often result in poor predictions. Some algorithms which perform well in prediction lack transparency, i.e., the discriminating features generated by these methods are not easily identifiable. In addition, the size of the sequence-based datasets presents practical challenges to most learning algorithms. Most sequence-based datasets contain millions or even billions of instances, for example, the genome-wide sequences of organisms in bioinformatics. At these sizes, classic learning algorithms often become prohibitively expensive, making scalability an important issue. Therefore, there is a need for an approach that can help find features/signals in complex sequences, offer meaningful discriminators, produce good predictions, and can scale well in time and space. This dissertation addresses the above issues by designing a comprehensive approach in the form of the Evolutionary Machine Learner (EML) framework. This framework can be employed on sequence-based datasets to generate explicit, human-recognizable features while solving the scalability issue. EML framework consists of a novel EA-based feature generation (EFG) algorithm for automatic feature construction. By modeling four complex sequencing problems in bioinformatics and generating meaningful, human-understandable features with comparable or better accuracy than the state of the art algorithms, the power and usefulness of the EFG algorithm is demonstrated. The EFG algorithm is also validated by applying it to time series classification problems showing the generic nature of the algorithm in finding the important discriminating patterns that assist in modeling sequence based data. EML framework addresses the scalability issue by means of a novel, parallel scalable machine learning algorithm (PSBML) based on spatially structured evolutionary algorithms. PSBML is validated on real-world "big data" classification problems for various properties of meta-learning, scalability and noise resilience using well known benchmark datasets. The PSBML algorithm is also proven theoretically to be a large margin classifier with linear scalability in training time and space, giving it a unique distinction among the existing large scale learning algorithms. Finally, the EML framework is validated on a large genome-wide bioinformatics classification problem and a large time series problem, showing that the combined algorithms achieve higher predictive performance, training time speed up, and the ability to produce human-understandable discriminating signals as features. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]

Red snapper, "Lutjanus campechanus," were sampled with hook and line at natural (n = 33) and artificial (n = 27) reef sites in the northern Gulf of Mexico from 2009-2011. Stomachs (n = 708) were extracted and their contents preserved for gut content analysis, and muscle tissue samples (n = 200) were dissected and frozen for stable isotope ratio-mass spectrometry (delta[superscript 13]C, delta[superscript 15]N, and delta[superscript 34]S) analysis. Forty-eight percent of sampled fish had identifiable prey which were grouped into 7 categories: fish, decapods, cephalopods, stomatopods, gastropods, zooplankton, and other invertebrates. A permutational multivariate analysis of variance (PERMANOVA) was computed to test the effect of fish size, habitat type, and the Deepwater Horizon Oil Spill (DHOS) on red snapper diet. Fish size (PERMANOVA, p = 0.021), and the DHOS (PERMANOVA, p = 0.001) were both significant effects in the model, but interactions between habitat type and the DHOS (PERMANOVA, p = 0.049) was also significant. Significant differences in diet among red snapper size categories were due to low trophic position prey items, such as pelagic zooplankton, being more abundant in the diet of larger (>500mm) red snapper, while red snapper in smaller size classes contained greater amounts of higher trophic level prey such decapods and fish. Although no significant difference existed in red snapper diet between habitat types, they consumed slightly higher amounts of decapods at artificial reefs (21.9% of total diet) than natural (14.8%). The significant DHOS effect was driven by an increase in fish and a decrease in zooplankton in red snapper diet following the spill; however, this effect cannot be interpreted independently given the significant interaction between habitat type and the DHOS effect. Zooplankton prey consumption declined and fish consumption increased for red snapper sampled at both artificial and natural reefs after the DHOS, but fish consumption increased to a greater extent at artificial reef sites following the spill. A significant increase in delta[superscript 15]N after the DHOS (ANOVA, p <0.001) also indicates red snapper sampled fed on higher trophic position prey following the spill, and lower delta[superscript 34]S values indicate an increase in benthic versus pelagic prey consumption following the DHOS (ANOVA, p <0.001). The ontogenetic shift in diet observed via gut content analysis was also apparent in the significant effect of fish size on delta[superscript 15]N (ANOVA, p = 0.015). Overall, results clearly indicate the DHOS affected red snapper diet and trophic position, and likely affected the abundance of prey resources as well. Study results also confirm the utility of stable isotope analysis to infer aspects of reef fish trophic ecology. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com/en-US/products/dissertations/individuals.shtml.]

Neurodegenerative diseases are pathologically characterized by the induction of gliotic states and the loss of specific neurons in the human brain. A precise ascertainment of cellular states altered and those types of neurons that are lost has not been made in Alzheimer’s and Parkinson’s disease, the two most common forms of brain degeneration. We employed new single-cell genomic technologies to better understand the cellular alterations that occur in both these diseases. In Parkinson’s disease, we developed a method to enrich midbrain dopaminergic neuronal nuclei from postmortem human samples. Using this strategy, we profiled thousands of these neurons to identify the diversity of these different types in the midbrain. Strikingly, we found one population, confined to the ventral tier of the pars compacta, was uniquely susceptible to Parkinson’s-associated cell loss. This same population was enriched for common variant heritable risk of the disease, suggesting cell-autonomous mechanisms underlie the genetics of PD. To better understand the cellular alterations in Alzheimer’s disease before death, we profiled individuals with suspected idiopathic normal pressure hydrocephalus (iNPH) with comorbid AD pathology. We identified two microglia populations, one increased and one decreased in abundance respectively, and confirmed these alterations in postmortem AD datasets. We further identified a single interneuron subpopulation residing in the uppermost layer of the neocortex as lost especially in the earliest stages of AD pathology. Finally, we find oligodendrocytes are an unrecognized contributor to beta-amyloid pathology. Measurement of beta-amyloid from stem cell-derived human oligodendrocytes confirmed that these cells are a significant producer of this important peptide. This body of work offers compelling evidence towards identifying the specific cellular state changes in association with two common brain diseases, offering new and testable hypotheses for future biological investigations into their pathogenesis.

Behavioral ecology is the study of how animal behavior evolves in the context of ecology, thus melding, by definition, investigations of how social, ecological, and evolutionary forces shape phenotypic variation within and across species. Framed thus, it is apparent that behavioral ecology also aims to cut across temporal scales and levels of biological organization, seeking to explain the long-term evolutionary trajectory of populations and species by understanding short-term interactions at the within-population level. In this dissertation, I make the case that paying attention to individuals’ natural history—where and how individual organisms live and whom and what they interact with, in natural conditions—can open avenues into studying the behavioral ecology of previously understudied organisms, and more importantly, recast our understanding of taxa we think we know well. In Chapters 1 and 2, I investigate the consequences of paying attention to individuals’ movement patterns for understanding the mating system of Anolis lizards (Dactyloidae). Chapter 1 is a historical review of Anolis social and reproductive behavior, in which I show that because individual variation in movement and social behavior was consistently ignored or downplayed in studies of Anolis mating systems, the conclusion that anoles are territorial and polygynous remains poorly supported by behavioral evidence. In Chapter 2, I examine movement behavior in a population of Anolis sagrei, and show that individuals’ movement behavior allows females to encounter and potentially mate with multiple males, which is counter to previous behavioral descriptions of Anolis mating systems but is consistent with growing genetic evidence for female multiple mating. The potential implications for sexual selection are discussed. In Chapter 3, I show that individual A. sagrei are specialized in habitat use, and that the degree of specialization among individuals in the sampled population in Gainesville, FL, is comparable to the degree of specialization across habitat specialist species in a community of Greater Antillean anoles. However, individual specialization in habitat use is not related to available habitat or to the morphological traits that explain habitat use specialization across species and populations of habitat use, suggesting that macroevolutionary patterns in this group are not simply microevolutionary patterns “writ large.” Finally, in Chapter 4, I quantify variation in display behavior, morphology, and habitat in the fan-throated lizard (Sitana and Sarada; Agamidae) species complex, and find that display behavior is partially but not completely associated with ornament elaboration in this group. Moreover, though habitat structure does not play a role in the maintenance of throat-fan and display variation, sexual selection very likely shaped the dramatic variation in ornament morphology in this group., Biology, Organismic and Evolutionary, behavior, lizard