{"id":{"repo_id":"ku","oai_identifier":"oai:kuscholarworks.ku.edu:1808/39485"},"canonical_url":"https://search.dev.ndltd.org/etd/ku/oai:kuscholarworks.ku.edu:1808/39485","repository":{"repo_id":"ku","name":"University of Kansas","base_url":"https://kuscholarworks.ku.edu/server/oai/request"},"display":{"title":"From Traditional Morphology to Genomes and Machine Learning: Navigating Taxonomic Uncertainties in Southeast Asian Agamids","abstract":"Classifying species, or the process of identifying, naming, and grouping organisms based on shared characteristics and evolutionary relationships, is critical in many aspects of biology. Without species classification, it would be difficult, if not impossible, for us to understand Earth’s biodiversity and the processes that generate it. However, classifying organisms into discrete units can be challenging, particularly when a historical species is named without designating an associated name-bearing type specimen(s) or when available names have been used irregularly in older, historical literature. These issues are exacerbated when a group is difficult to diagnose because phenotypic distinctiveness and genetic divergence do not align, as in the case of cryptic species, or when a group lacks considerable genetic divergence but exhibits variable morphology, as in polymorphic species. My dissertation provides an empirical example of delimiting species boundaries, with nomenclatural issues compounded by unclear signals of genetic and morphological variation and demonstrates how previously ambiguous taxonomies, muddled by historical confusion, can be navigated by careful integration of all available data and use of modern tools.","abstract_html":"Classifying species, or the process of identifying, naming, and grouping organisms based on shared characteristics and evolutionary relationships, is critical in many aspects of biology. Without species classification, it would be difficult, if not impossible, for us to understand Earth’s biodiversity and the processes that generate it. However, classifying organisms into discrete units can be challenging, particularly when a historical species is named without designating an associated name-bearing type specimen(s) or when available names have been used irregularly in older, historical literature. These issues are exacerbated when a group is difficult to diagnose because phenotypic distinctiveness and genetic divergence do not align, as in the case of cryptic species, or when a group lacks considerable genetic divergence but exhibits variable morphology, as in polymorphic species. My dissertation provides an empirical example of delimiting species boundaries, with nomenclatural issues compounded by unclear signals of genetic and morphological variation and demonstrates how previously ambiguous taxonomies, muddled by historical confusion, can be navigated by careful integration of all available data and use of modern tools.","abstract_has_math":false,"creators":["Supsup, Christian E."],"institution":"University of Kansas","degree_name":"Ph.D.","degree_level":null,"degree_discipline":"Ecology & Evolutionary Biology","degree_department":null,"school":null,"contributors":[],"advisors":["Brown, Rafe M."],"committee_chairs":[],"committee_members":[],"year":2026,"date_issued":"2026-05-31","date_published":"2026-05-31","updated_at":"2026-07-24T02:45:07Z","subjects":["genome","neural network","phylogenetics","population genetics","species delimitation","taxonomy"],"languages":["en"],"rights":[],"rights_urls":[],"identifier_entries":[{"key":"dc:identifier.other","label":"Dc Identifier Other","values":["https://www.proquest.com/LegacyDocView/DISSNUM/32697970"],"render_values":[{"text":"https://www.proquest.com/LegacyDocView/DISSNUM/32697970","href":"https://www.proquest.com/LegacyDocView/DISSNUM/32697970","code":true}]}]},"links":{"outbound_url":"https://hdl.handle.net/1808/39485","outbound_label":"Handle","outbound_source":"dc:identifier.uri"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor.advisor","label":"Advisor","values":["Brown, Rafe M."]},{"key":"dc:creator","label":"Author","values":["Supsup, Christian E."]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.accessioned","label":"Dc Date Accessioned","values":["2026-07-15T22:20:23Z"]},{"key":"dc:date.available","label":"Dc Date Available","values":["2026-07-15T22:20:23Z"]},{"key":"dc:date.issued","label":"Date","values":["2026-05-31"]},{"key":"dc:publisher","label":"Institution","values":["University of Kansas"]},{"key":"dc:type","label":"Dc Type","values":["Dissertation"]},{"key":"thesis:degree_discipline","label":"Discipline","values":["Ecology & Evolutionary Biology"]},{"key":"thesis:degree_name","label":"Degree Name","values":["Ph.D."]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["genome","neural network","phylogenetics","population genetics","species delimitation","taxonomy"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language.iso","label":"Language (ISO)","values":["en"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.other","label":"Dc Identifier Other","values":["https://www.proquest.com/LegacyDocView/DISSNUM/32697970"]},{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://hdl.handle.net/1808/39485"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["Classifying species, or the process of identifying, naming, and grouping organisms based on shared characteristics and evolutionary relationships, is critical in many aspects of biology. Without species classification, it would be difficult, if not impossible, for us to understand Earth’s biodiversity and the processes that generate it. However, classifying organisms into discrete units can be challenging, particularly when a historical species is named without designating an associated name-bearing type specimen(s) or when available names have been used irregularly in older, historical literature. These issues are exacerbated when a group is difficult to diagnose because phenotypic distinctiveness and genetic divergence do not align, as in the case of cryptic species, or when a group lacks considerable genetic divergence but exhibits variable morphology, as in polymorphic species. My dissertation provides an empirical example of delimiting species boundaries, with nomenclatural issues compounded by unclear signals of genetic and morphological variation and demonstrates how previously ambiguous taxonomies, muddled by historical confusion, can be navigated by careful integration of all available data and use of modern tools.","In Chapter I, I investigated the Green Crested Lizard (genus Bronchocela) of the family Agamidae from the Philippine Archipelago and neighboring Southeast Asian regions to resolve long-standing taxonomic issues within the group that stem from a lack of type specimens and locality information in the original description, along with highly similar morphology that can confuse taxonomists. I conducted an archipelago-wide analysis of genetic (mitochondrial DNA) data to reconstruct the first comprehensive phylogenetic relationships of the Philippine Bronchocela. Additionally, I examined hundreds of museum specimens to screen for morphological variation and trace the possible origin of the type species (B. cristatella). I found that the Philippine Bronchocela population comprises multiple distinct lineages, leading to a taxonomic revision of the genus in the archipelago that includes the description of two new species and the resurrection of a previously synonymized name. This work also restricted the type locality of B. cristatella to the Indonesian islands of Java and Sumatra (southern region), thereby paving the way for future taxonomic efforts to stabilize species-level taxonomy within the genus.","In Chapter II, I examined the Philippine Angle-headed Lizards (genus Gonocephalus), also from the family Agamidae, which share similar taxonomic issues with Philippine Bronchocela, but to a lesser extent. Taxonomic confusion within the Philippine Gonocephalus originated from insufficient locality details for type specimens, provided merely as “The Philippines,” and from the irregular use of available names in the literature, due to inconsistent morphological diagnostic characters, resulting in uncertain species boundaries. This taxonomic problem persisted for more than a century until recent multi-locus molecular work by Welton et al. (2017), which revealed at least 12 putative lineages from the archipelago. However, the relation of the 12 lineages to the currently known three species of Philippine Gonocephalus is still unclear due to a lack of morphological assessment and examination of name-bearing type specimens. I conducted a taxonomic reevaluation of the group by testing the hypothesized 12 lineages using newly generated genomic and morphological data. In my phylogenomic analysis, I recovered at least 12 lineages, similar to those reported by Welton et al., embedded within two highly supported clades. Lineages within the two main clades exhibited considerable topological discordance, characterized by minimal genomic differentiation and high levels of shared ancestry, but such discordance was not observed between clades, which were, instead, strongly supported as reciprocally monophyletic. Morphological assessment supported two diagnosable groups, corresponding to the two well-supported clades. Overall, my extensive reassessment of the Philippine Gonocelaphus supports the recognition of only two distinct species, not three. Examination of name-bearing types clarified the appropriate nomenclatural assignment to two clades; I recognize G. sophiae in the northern and central Philippines, and G. semperi in the southern portions of the archipelago, and I place G. interruptus in synonymy with G. semperi (the name with chronological precedence).","In Chapter III, I explored several machine learning methods as we continue to seek better approaches for delimiting species boundaries. Species delimitation ideally integrates multiple independent data types (e.g., genetic, morphology, environment, etc.) and analyzes them independently or jointly, as no single type of data captures the complexity of a species. However, in most cases, as in Chapters I and II, genetic and morphological data were considered independently, and when congruent results were found across data types, confidence in species delimitation increased. This raises the question of whether independent analysis of each data type is truly integrative. Since recent developments in species delimitation analysis are mostly designed for independent analysis of data types, I tested three unsupervised neural network-based machine learning methods (Improved Deep Embedding Clustering, Variational Deep Embedding, and Self-organizing Maps) to analyze multiple independent data types for joint species delimitation (JSD). To test these methods for JSD, I used both simulated and empirical datasets, which included genomic, morphological, and environmental data. I first asked whether machine learning methods are appropriate for performing JSD. I found that using machine learning can handle multiple data types and infer the expected species cluster without over-splitting, given the optimal parameter configuration for the learning process. Then, I asked whether combining data types is necessary for a truly integrative species delimitation. My results indicate that combining data types yields more consistent species inference when a common signal of variation is shared across data types, however, this becomes ineffective when the shared signal is weak or absent, leading to inconsistent species inference. I also highlight several challenges of using neural network methods for species delimitation, including hyperparameter tuning and output selection, amidst hype and their growing popularity and applications to species delimitation.Together, my three chapters demonstrate that species classification remains challenging, especially when faced with taxonomic uncertainties arising from a lack of critical information (e.g., type specimens, precise locality) and insufficient data necessary to delineate species boundaries. This dissertation was able to clarify and delimit species boundaries among the two groups of agamid lizards through sustained efforts by many individuals and institutions that enabled the collection of many specimens used to understand genetic and morphological variation. However, many groups with taxonomic issues awaiting resolution, particularly those considered rare and thus difficult to observe or collect, still pose challenges that need careful consideration. Furthermore, with the use of more comprehensive data (e.g., genomic data types, robust and equitable geographic sampling) and the development of modern tools, species classification may not necessarily become easier as we continue to analyze and unpack the complex details of the speciation spectrum. As demonstrated here, what is most important is the use of all available data and tools to provide evidence to support or overturn species hypotheses."]},{"key":"dc:title","label":"Title","values":["From Traditional Morphology to Genomes and Machine Learning: Navigating Taxonomic Uncertainties in Southeast Asian Agamids"]}]}],"canonical_facts":{"dc:contributor.advisor":["Brown, Rafe M."],"dc:creator":["Supsup, Christian E."],"dc:date.accessioned":["2026-07-15T22:20:23Z"],"dc:date.available":["2026-07-15T22:20:23Z"],"dc:date.issued":["2026-05-31"],"dc:description.abstract":["Classifying species, or the process of identifying, naming, and grouping organisms based on shared characteristics and evolutionary relationships, is critical in many aspects of biology. Without species classification, it would be difficult, if not impossible, for us to understand Earth’s biodiversity and the processes that generate it. However, classifying organisms into discrete units can be challenging, particularly when a historical species is named without designating an associated name-bearing type specimen(s) or when available names have been used irregularly in older, historical literature. These issues are exacerbated when a group is difficult to diagnose because phenotypic distinctiveness and genetic divergence do not align, as in the case of cryptic species, or when a group lacks considerable genetic divergence but exhibits variable morphology, as in polymorphic species. My dissertation provides an empirical example of delimiting species boundaries, with nomenclatural issues compounded by unclear signals of genetic and morphological variation and demonstrates how previously ambiguous taxonomies, muddled by historical confusion, can be navigated by careful integration of all available data and use of modern tools.","In Chapter I, I investigated the Green Crested Lizard (genus Bronchocela) of the family Agamidae from the Philippine Archipelago and neighboring Southeast Asian regions to resolve long-standing taxonomic issues within the group that stem from a lack of type specimens and locality information in the original description, along with highly similar morphology that can confuse taxonomists. I conducted an archipelago-wide analysis of genetic (mitochondrial DNA) data to reconstruct the first comprehensive phylogenetic relationships of the Philippine Bronchocela. Additionally, I examined hundreds of museum specimens to screen for morphological variation and trace the possible origin of the type species (B. cristatella). I found that the Philippine Bronchocela population comprises multiple distinct lineages, leading to a taxonomic revision of the genus in the archipelago that includes the description of two new species and the resurrection of a previously synonymized name. This work also restricted the type locality of B. cristatella to the Indonesian islands of Java and Sumatra (southern region), thereby paving the way for future taxonomic efforts to stabilize species-level taxonomy within the genus.","In Chapter II, I examined the Philippine Angle-headed Lizards (genus Gonocephalus), also from the family Agamidae, which share similar taxonomic issues with Philippine Bronchocela, but to a lesser extent. Taxonomic confusion within the Philippine Gonocephalus originated from insufficient locality details for type specimens, provided merely as “The Philippines,” and from the irregular use of available names in the literature, due to inconsistent morphological diagnostic characters, resulting in uncertain species boundaries. This taxonomic problem persisted for more than a century until recent multi-locus molecular work by Welton et al. (2017), which revealed at least 12 putative lineages from the archipelago. However, the relation of the 12 lineages to the currently known three species of Philippine Gonocephalus is still unclear due to a lack of morphological assessment and examination of name-bearing type specimens. I conducted a taxonomic reevaluation of the group by testing the hypothesized 12 lineages using newly generated genomic and morphological data. In my phylogenomic analysis, I recovered at least 12 lineages, similar to those reported by Welton et al., embedded within two highly supported clades. Lineages within the two main clades exhibited considerable topological discordance, characterized by minimal genomic differentiation and high levels of shared ancestry, but such discordance was not observed between clades, which were, instead, strongly supported as reciprocally monophyletic. Morphological assessment supported two diagnosable groups, corresponding to the two well-supported clades. Overall, my extensive reassessment of the Philippine Gonocelaphus supports the recognition of only two distinct species, not three. Examination of name-bearing types clarified the appropriate nomenclatural assignment to two clades; I recognize G. sophiae in the northern and central Philippines, and G. semperi in the southern portions of the archipelago, and I place G. interruptus in synonymy with G. semperi (the name with chronological precedence).","In Chapter III, I explored several machine learning methods as we continue to seek better approaches for delimiting species boundaries. Species delimitation ideally integrates multiple independent data types (e.g., genetic, morphology, environment, etc.) and analyzes them independently or jointly, as no single type of data captures the complexity of a species. However, in most cases, as in Chapters I and II, genetic and morphological data were considered independently, and when congruent results were found across data types, confidence in species delimitation increased. This raises the question of whether independent analysis of each data type is truly integrative. Since recent developments in species delimitation analysis are mostly designed for independent analysis of data types, I tested three unsupervised neural network-based machine learning methods (Improved Deep Embedding Clustering, Variational Deep Embedding, and Self-organizing Maps) to analyze multiple independent data types for joint species delimitation (JSD). To test these methods for JSD, I used both simulated and empirical datasets, which included genomic, morphological, and environmental data. I first asked whether machine learning methods are appropriate for performing JSD. I found that using machine learning can handle multiple data types and infer the expected species cluster without over-splitting, given the optimal parameter configuration for the learning process. Then, I asked whether combining data types is necessary for a truly integrative species delimitation. My results indicate that combining data types yields more consistent species inference when a common signal of variation is shared across data types, however, this becomes ineffective when the shared signal is weak or absent, leading to inconsistent species inference. I also highlight several challenges of using neural network methods for species delimitation, including hyperparameter tuning and output selection, amidst hype and their growing popularity and applications to species delimitation.Together, my three chapters demonstrate that species classification remains challenging, especially when faced with taxonomic uncertainties arising from a lack of critical information (e.g., type specimens, precise locality) and insufficient data necessary to delineate species boundaries. This dissertation was able to clarify and delimit species boundaries among the two groups of agamid lizards through sustained efforts by many individuals and institutions that enabled the collection of many specimens used to understand genetic and morphological variation. However, many groups with taxonomic issues awaiting resolution, particularly those considered rare and thus difficult to observe or collect, still pose challenges that need careful consideration. Furthermore, with the use of more comprehensive data (e.g., genomic data types, robust and equitable geographic sampling) and the development of modern tools, species classification may not necessarily become easier as we continue to analyze and unpack the complex details of the speciation spectrum. As demonstrated here, what is most important is the use of all available data and tools to provide evidence to support or overturn species hypotheses."],"dc:identifier.other":["https://www.proquest.com/LegacyDocView/DISSNUM/32697970"],"dc:identifier.uri":["https://hdl.handle.net/1808/39485"],"dc:language.iso":["en"],"dc:publisher":["University of Kansas"],"dc:subject":["genome","neural network","phylogenetics","population genetics","species delimitation","taxonomy"],"dc:title":["From Traditional Morphology to Genomes and Machine Learning: Navigating Taxonomic Uncertainties in Southeast Asian Agamids"],"dc:type":["Dissertation"],"thesis:degree_discipline":["Ecology & Evolutionary Biology"],"thesis:degree_name":["Ph.D."]},"updated_at":"2026-07-24T02:45:07Z"}