{"id":{"repo_id":"edinburgh","oai_identifier":"oai:era.ed.ac.uk:1842/44853"},"canonical_url":"https://search.dev.ndltd.org/etd/edinburgh/oai:era.ed.ac.uk:1842/44853","repository":{"repo_id":"edinburgh","name":"University of Edinburgh","base_url":"https://era.ed.ac.uk/server/oai/request"},"display":{"title":"Genomic and metabolomic interrogation of body composition and insulin resistance endotypes","abstract":"Metabolic syndrome and related cardiometabolic disorders are common yet highly heterogeneous, reflecting different ways in which the complex metabolic interplay among insulin-sensitive tissues can go awry. Current clinical classification remains imprecise, however, relying on broad constructs (obesity, type 2 diabetes, BMI thresholds) that do not reliably distinguish underlying mechanisms. One way to inform research into underpinning mechanisms is to anchor it on rare and severe disorders of known cause. In the case of this study this means disorders featuring both severe metabolic derangement and perturbation of body composition. Such rare monogenic disorders reveal causal pathways at the extremes, while population-scale analyses, i.e. genome/exome wide association studies (GWAS/ExWAS), identify common variation shaping risk of more prevalent forms of disease. Healthy skeletal muscle and adipose tissue support physical function and cardiometabolic resilience across the life course. In the general population age-related lean-mass loss and/or disrupted adipose homeostasis are associated with frailty/sarcopenia, insulin resistance (IR), and adverse outcomes. Sentinel monogenic disorders demonstrate that such disproportion between (muscle/adipose) compartments can have a single unifying cause. In the case of increased muscle and reduced adipose tissue this can correlate with either pathological hypofunction i.e. lipodystrophy (LD) or, more rarely, enhanced function and resilience (myostatin deficiency). It was hypothesised that in the general population commoner genetic variants influencing the relative proportion of muscle and fat may also be found, illuminating mechanisms preserving healthy body composition and maintaining metabolic resilience. This thesis first leverages large-scale population genetics, using GWAS/ExWAS to interrogate body-composition phenotypes grounded in observation of paradigmatic monogenic disorders. First, bi-trait analyses of whole body fat mass and fat-free mass normalised by height (FMI and FFMI) were performed using bioelectrical impedance measures from UK Biobank. Variant-level association analyses were conducted for each trait, followed by cross-phenotype tests to identify loci with diverging effects on FFMI and FMI, with locus-level filtering applied to retain signals supported across both traits. This identified 6 variants in men, 5 in women, and 14 in sex-combined analyses. These variants were examined across a panel of cardiometabolic and body-composition phenotypes, and association profiles were compared with rare-variant gene-burden scores in putative mediating genes. Across loci and clusters, PLCE1 and IRS1 showed the most consistent cross-phenotype evidence in the provided analyses, alongside additional supported candidates including PPARG (notably for lipodystrophy-like metabolic correlates) and MTMR11 (for height/anthropometry-heavy signatures), reflecting heterogeneity in the downstream consequences of an FFMI-increasing/FMI-decreasing pattern. A complementary exome-based strategy used gene-based rare-variant burden testing in UK Biobank whole-exome sequencing to analyse association with the composite contrast (FFMI - FMI), and then decomposed significant composite associations into their component FFMI and FMI effects for interpretability. Under the combined functional mask (loss-of-function plus protein-altering missense; alternate allele frequency ≤ 0.001), 25 genes met exome-wide FDR significance for FFMI − FMI while also showing opposite-direction effects on FFMI and FMI. The strongest signals highlighted an extracellular matrix/remodelling theme (including multiple ADAMTS/ADAMTS-like family genes), alongside exemplars with mechanistically informative profiles such as ATP2A1 (lean-mass-aligned with supporting biochemical associations) and GH1 (opposite-direction effects consistent with growth-hormone biology). PLCE1 appeared as a cross-chapter bridge, being supported by both common-variant locus-based analyses and rare-variant gene-based evidence, while emphasising that interpretation and translation should consider known Mendelian disease context for some prioritised genes. The second body of work in this thesis uses metabolomics to assess endophenotypes and disease biomarkers in a population of people with severe IR of known aetiology, mediated by either primary insulin signalling defects or lipodystrophy. Because prognosis and management of these disorders differ sharply, and because diagnosis is often severely delayed, new biomarkers would be valuable. A further motivation was to deepen mechanistic understanding of the differences in these IR subphenotypes to enhance understanding of the common metabolic syndrome. Two independent datasets (Cambridge and NIH) including metabolomic data generated from the same technical platform were analysed. The Cambridge dataset comprised 87 females (LD n = 65; insulin signalling group n = 22). The NIH dataset comprised 59 individuals (LD n = 30; insulin signalling group n = 29). Each cohort was processed separately to avoid cross-cohort artefacts. Metabolite abundances were log-transformed (as provided by the vendor) and robustly scaled, with quality-control checks demonstrating near-identical high-dimensional correlation structure after processing. Highly correlated metabolites were collapsed using medoid reduction. Feature selection used a genetic algorithm to identify metabolite subsets that best separated clinically assigned labels (lipodystrophy vs insulin receptoropathy). Performance was evaluated using cosine-distance-based clustering/classification, summarised by adjusted Rand index (ARI), silhouette score (SS), and misclassification counts, with stability across independent island searches assessed using Jaccard similarity. Across cohorts, the strongest discriminatory structure was carried by lipid and amino-acid domains, while the combined non-lipid/non-amino-acid panel consistently underperformed. In Cambridge, separation was strong for lipids (115-metabolite consensus set; ARI = 0.810; SS = 0.085; 4 misclassifications) and particularly strong for amino acids (138-metabolite consensus set; ARI = 0.951; SS = 0.124; 1 misclassification), whereas the combined panel did not yield meaningful two-group separation (55-metabolite consensus set; ARI close to 0). In NIH, consensus signatures achieved good separation for lipids (144-metabolite consensus set; ARI = 0.804; SS = 0.102; 3 misclassifications) and amino acids (106-metabolite consensus set; ARI = 0.743; SS = 0.177; 4 misclassifications), but only modest separation for the combined panel (66-metabolite consensus set; ARI = 0.263). Complex lipid panels provided some of the cleanest endotype discrimination in both cohorts (Cambridge: 178-metabolite consensus set; ARI = 0.952; SS = 0.231; 1 misclassification; NIH: 171-metabolite consensus set; ARI = 0.804; SS = 0.286; 3 misclassifications), supporting a prominent neutral-lipid and phospholipid-remodelling signal and a recurring ether/plasmalogen axis. Findings across both cohorts are consistent with differences in PUFA remodelling, mitochondrial fatty-acid handling, glycerophospholipid metabolism, and amino-acid utilisation/acetylation in LD relative to primary insulin signalling defects, including perturbations in BCAA flux and tryptophan–kynurenine/indole metabolism. The studies in this thesis use complementary strategies to refine cardiometabolic heterogeneity into mechanistically informative endotypes, first identifying common-variant and gene-level contributors to lean-fat partitioning, nominating loci and genes plausibly contributing to healthier tissue distribution and physical resilience across ageing, and second refining metabolomic signatures discriminating lipodystrophic and primary insulin resistance. This advances a framework for moving beyond broad clinical labels towards biologically grounded stratification, while prioritising testable pathways and candidate targets for mechanistic follow-up.","abstract_html":"Metabolic syndrome and related cardiometabolic disorders are common yet highly heterogeneous, reflecting different ways in which the complex metabolic interplay among insulin-sensitive tissues can go awry. Current clinical classification remains imprecise, however, relying on broad constructs (obesity, type 2 diabetes, BMI thresholds) that do not reliably distinguish underlying mechanisms. One way to inform research into underpinning mechanisms is to anchor it on rare and severe disorders of known cause. In the case of this study this means disorders featuring both severe metabolic derangement and perturbation of body composition. Such rare monogenic disorders reveal causal pathways at the extremes, while population-scale analyses, i.e. genome/exome wide association studies (GWAS/ExWAS), identify common variation shaping risk of more prevalent forms of disease. Healthy skeletal muscle and adipose tissue support physical function and cardiometabolic resilience across the life course. In the general population age-related lean-mass loss and/or disrupted adipose homeostasis are associated with frailty/sarcopenia, insulin resistance (IR), and adverse outcomes. Sentinel monogenic disorders demonstrate that such disproportion between (muscle/adipose) compartments can have a single unifying cause. In the case of increased muscle and reduced adipose tissue this can correlate with either pathological hypofunction i.e. lipodystrophy (LD) or, more rarely, enhanced function and resilience (myostatin deficiency). It was hypothesised that in the general population commoner genetic variants influencing the relative proportion of muscle and fat may also be found, illuminating mechanisms preserving healthy body composition and maintaining metabolic resilience. This thesis first leverages large-scale population genetics, using GWAS/ExWAS to interrogate body-composition phenotypes grounded in observation of paradigmatic monogenic disorders. First, bi-trait analyses of whole body fat mass and fat-free mass normalised by height (FMI and FFMI) were performed using bioelectrical impedance measures from UK Biobank. Variant-level association analyses were conducted for each trait, followed by cross-phenotype tests to identify loci with diverging effects on FFMI and FMI, with locus-level filtering applied to retain signals supported across both traits. This identified 6 variants in men, 5 in women, and 14 in sex-combined analyses. These variants were examined across a panel of cardiometabolic and body-composition phenotypes, and association profiles were compared with rare-variant gene-burden scores in putative mediating genes. Across loci and clusters, PLCE1 and IRS1 showed the most consistent cross-phenotype evidence in the provided analyses, alongside additional supported candidates including PPARG (notably for lipodystrophy-like metabolic correlates) and MTMR11 (for height/anthropometry-heavy signatures), reflecting heterogeneity in the downstream consequences of an FFMI-increasing/FMI-decreasing pattern. A complementary exome-based strategy used gene-based rare-variant burden testing in UK Biobank whole-exome sequencing to analyse association with the composite contrast (FFMI - FMI), and then decomposed significant composite associations into their component FFMI and FMI effects for interpretability. Under the combined functional mask (loss-of-function plus protein-altering missense; alternate allele frequency ≤ 0.001), 25 genes met exome-wide FDR significance for FFMI − FMI while also showing opposite-direction effects on FFMI and FMI. The strongest signals highlighted an extracellular matrix/remodelling theme (including multiple ADAMTS/ADAMTS-like family genes), alongside exemplars with mechanistically informative profiles such as ATP2A1 (lean-mass-aligned with supporting biochemical associations) and GH1 (opposite-direction effects consistent with growth-hormone biology). PLCE1 appeared as a cross-chapter bridge, being supported by both common-variant locus-based analyses and rare-variant gene-based evidence, while emphasising that interpretation and translation should consider known Mendelian disease context for some prioritised genes. The second body of work in this thesis uses metabolomics to assess endophenotypes and disease biomarkers in a population of people with severe IR of known aetiology, mediated by either primary insulin signalling defects or lipodystrophy. Because prognosis and management of these disorders differ sharply, and because diagnosis is often severely delayed, new biomarkers would be valuable. A further motivation was to deepen mechanistic understanding of the differences in these IR subphenotypes to enhance understanding of the common metabolic syndrome. Two independent datasets (Cambridge and NIH) including metabolomic data generated from the same technical platform were analysed. The Cambridge dataset comprised 87 females (LD n = 65; insulin signalling group n = 22). The NIH dataset comprised 59 individuals (LD n = 30; insulin signalling group n = 29). Each cohort was processed separately to avoid cross-cohort artefacts. Metabolite abundances were log-transformed (as provided by the vendor) and robustly scaled, with quality-control checks demonstrating near-identical high-dimensional correlation structure after processing. Highly correlated metabolites were collapsed using medoid reduction. Feature selection used a genetic algorithm to identify metabolite subsets that best separated clinically assigned labels (lipodystrophy vs insulin receptoropathy). Performance was evaluated using cosine-distance-based clustering/classification, summarised by adjusted Rand index (ARI), silhouette score (SS), and misclassification counts, with stability across independent island searches assessed using Jaccard similarity. Across cohorts, the strongest discriminatory structure was carried by lipid and amino-acid domains, while the combined non-lipid/non-amino-acid panel consistently underperformed. In Cambridge, separation was strong for lipids (115-metabolite consensus set; ARI = 0.810; SS = 0.085; 4 misclassifications) and particularly strong for amino acids (138-metabolite consensus set; ARI = 0.951; SS = 0.124; 1 misclassification), whereas the combined panel did not yield meaningful two-group separation (55-metabolite consensus set; ARI close to 0). In NIH, consensus signatures achieved good separation for lipids (144-metabolite consensus set; ARI = 0.804; SS = 0.102; 3 misclassifications) and amino acids (106-metabolite consensus set; ARI = 0.743; SS = 0.177; 4 misclassifications), but only modest separation for the combined panel (66-metabolite consensus set; ARI = 0.263). Complex lipid panels provided some of the cleanest endotype discrimination in both cohorts (Cambridge: 178-metabolite consensus set; ARI = 0.952; SS = 0.231; 1 misclassification; NIH: 171-metabolite consensus set; ARI = 0.804; SS = 0.286; 3 misclassifications), supporting a prominent neutral-lipid and phospholipid-remodelling signal and a recurring ether/plasmalogen axis. Findings across both cohorts are consistent with differences in PUFA remodelling, mitochondrial fatty-acid handling, glycerophospholipid metabolism, and amino-acid utilisation/acetylation in LD relative to primary insulin signalling defects, including perturbations in BCAA flux and tryptophan–kynurenine/indole metabolism. The studies in this thesis use complementary strategies to refine cardiometabolic heterogeneity into mechanistically informative endotypes, first identifying common-variant and gene-level contributors to lean-fat partitioning, nominating loci and genes plausibly contributing to healthier tissue distribution and physical resilience across ageing, and second refining metabolomic signatures discriminating lipodystrophic and primary insulin resistance. This advances a framework for moving beyond broad clinical labels towards biologically grounded stratification, while prioritising testable pathways and candidate targets for mechanistic follow-up.","abstract_has_math":false,"creators":["D'Silva, Sheldon"],"institution":null,"degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":null,"school":null,"contributors":[],"advisors":["Semple, Robert","Cawthorn, William","Loos, Ruth"],"committee_chairs":[],"committee_members":[],"year":2026,"date_issued":"2026-06-30","date_published":"2026-06-30","updated_at":"2026-07-24T02:14:14Z","subjects":["Genomics","GWAS","ExWAS","Metabolomics","Insulin Resistance (IR)","IR","Body Composition","Cardiometabolic Disorders"],"languages":["en"],"rights":[],"rights_urls":[],"identifier_entries":[{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://doi.org/10.7488/era/7367"],"render_values":[{"text":"https://doi.org/10.7488/era/7367","href":"https://doi.org/10.7488/era/7367","code":true}]}]},"links":{"outbound_url":"https://era.ed.ac.uk/handle/1842/44853","outbound_label":"Repository record","outbound_source":"dc:identifier.uri"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor.advisor","label":"Advisor","values":["Semple, Robert","Cawthorn, William","Loos, Ruth"]},{"key":"dc:creator","label":"Author","values":["D'Silva, Sheldon"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.accessioned","label":"Dc Date Accessioned","values":["2026-06-30T15:55:04Z"]},{"key":"dc:date.issued","label":"Date","values":["2026-06-30"]},{"key":"dc:type","label":"Dc Type","values":["Thesis"]},{"key":"dc:type.qualificationlevel","label":"Dc Type Qualificationlevel","values":["Doctoral"]},{"key":"dc:type.qualificationname","label":"Dc Type Qualificationname","values":["PhD Doctor of Philosophy"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Genomics","GWAS","ExWAS","Metabolomics","Insulin Resistance (IR)","IR","Body Composition","Cardiometabolic Disorders"]}]},{"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.uri","label":"Identifier URI","values":["https://era.ed.ac.uk/handle/1842/44853","https://doi.org/10.7488/era/7367"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["Metabolic syndrome and related cardiometabolic disorders are common yet highly heterogeneous, reflecting different ways in which the complex metabolic interplay among insulin-sensitive tissues can go awry. Current clinical classification remains imprecise, however, relying on broad constructs (obesity, type 2 diabetes, BMI thresholds) that do not reliably distinguish underlying mechanisms. One way to inform research into underpinning mechanisms is to anchor it on rare and severe disorders of known cause. In the case of this study this means disorders featuring both severe metabolic derangement and perturbation of body composition. Such rare monogenic disorders reveal causal pathways at the extremes, while population-scale analyses, i.e. genome/exome wide association studies (GWAS/ExWAS), identify common variation shaping risk of more prevalent forms of disease. Healthy skeletal muscle and adipose tissue support physical function and cardiometabolic resilience across the life course. In the general population age-related lean-mass loss and/or disrupted adipose homeostasis are associated with frailty/sarcopenia, insulin resistance (IR), and adverse outcomes. Sentinel monogenic disorders demonstrate that such disproportion between (muscle/adipose) compartments can have a single unifying cause. In the case of increased muscle and reduced adipose tissue this can correlate with either pathological hypofunction i.e. lipodystrophy (LD) or, more rarely, enhanced function and resilience (myostatin deficiency). It was hypothesised that in the general population commoner genetic variants influencing the relative proportion of muscle and fat may also be found, illuminating mechanisms preserving healthy body composition and maintaining metabolic resilience. This thesis first leverages large-scale population genetics, using GWAS/ExWAS to interrogate body-composition phenotypes grounded in observation of paradigmatic monogenic disorders. First, bi-trait analyses of whole body fat mass and fat-free mass normalised by height (FMI and FFMI) were performed using bioelectrical impedance measures from UK Biobank. Variant-level association analyses were conducted for each trait, followed by cross-phenotype tests to identify loci with diverging effects on FFMI and FMI, with locus-level filtering applied to retain signals supported across both traits. This identified 6 variants in men, 5 in women, and 14 in sex-combined analyses. These variants were examined across a panel of cardiometabolic and body-composition phenotypes, and association profiles were compared with rare-variant gene-burden scores in putative mediating genes. Across loci and clusters, PLCE1 and IRS1 showed the most consistent cross-phenotype evidence in the provided analyses, alongside additional supported candidates including PPARG (notably for lipodystrophy-like metabolic correlates) and MTMR11 (for height/anthropometry-heavy signatures), reflecting heterogeneity in the downstream consequences of an FFMI-increasing/FMI-decreasing pattern. A complementary exome-based strategy used gene-based rare-variant burden testing in UK Biobank whole-exome sequencing to analyse association with the composite contrast (FFMI - FMI), and then decomposed significant composite associations into their component FFMI and FMI effects for interpretability. Under the combined functional mask (loss-of-function plus protein-altering missense; alternate allele frequency ≤ 0.001), 25 genes met exome-wide FDR significance for FFMI − FMI while also showing opposite-direction effects on FFMI and FMI. The strongest signals highlighted an extracellular matrix/remodelling theme (including multiple ADAMTS/ADAMTS-like family genes), alongside exemplars with mechanistically informative profiles such as ATP2A1 (lean-mass-aligned with supporting biochemical associations) and GH1 (opposite-direction effects consistent with growth-hormone biology). PLCE1 appeared as a cross-chapter bridge, being supported by both common-variant locus-based analyses and rare-variant gene-based evidence, while emphasising that interpretation and translation should consider known Mendelian disease context for some prioritised genes. The second body of work in this thesis uses metabolomics to assess endophenotypes and disease biomarkers in a population of people with severe IR of known aetiology, mediated by either primary insulin signalling defects or lipodystrophy. Because prognosis and management of these disorders differ sharply, and because diagnosis is often severely delayed, new biomarkers would be valuable. A further motivation was to deepen mechanistic understanding of the differences in these IR subphenotypes to enhance understanding of the common metabolic syndrome. Two independent datasets (Cambridge and NIH) including metabolomic data generated from the same technical platform were analysed. The Cambridge dataset comprised 87 females (LD n = 65; insulin signalling group n = 22). The NIH dataset comprised 59 individuals (LD n = 30; insulin signalling group n = 29). Each cohort was processed separately to avoid cross-cohort artefacts. Metabolite abundances were log-transformed (as provided by the vendor) and robustly scaled, with quality-control checks demonstrating near-identical high-dimensional correlation structure after processing. Highly correlated metabolites were collapsed using medoid reduction. Feature selection used a genetic algorithm to identify metabolite subsets that best separated clinically assigned labels (lipodystrophy vs insulin receptoropathy). Performance was evaluated using cosine-distance-based clustering/classification, summarised by adjusted Rand index (ARI), silhouette score (SS), and misclassification counts, with stability across independent island searches assessed using Jaccard similarity. Across cohorts, the strongest discriminatory structure was carried by lipid and amino-acid domains, while the combined non-lipid/non-amino-acid panel consistently underperformed. In Cambridge, separation was strong for lipids (115-metabolite consensus set; ARI = 0.810; SS = 0.085; 4 misclassifications) and particularly strong for amino acids (138-metabolite consensus set; ARI = 0.951; SS = 0.124; 1 misclassification), whereas the combined panel did not yield meaningful two-group separation (55-metabolite consensus set; ARI close to 0). In NIH, consensus signatures achieved good separation for lipids (144-metabolite consensus set; ARI = 0.804; SS = 0.102; 3 misclassifications) and amino acids (106-metabolite consensus set; ARI = 0.743; SS = 0.177; 4 misclassifications), but only modest separation for the combined panel (66-metabolite consensus set; ARI = 0.263). Complex lipid panels provided some of the cleanest endotype discrimination in both cohorts (Cambridge: 178-metabolite consensus set; ARI = 0.952; SS = 0.231; 1 misclassification; NIH: 171-metabolite consensus set; ARI = 0.804; SS = 0.286; 3 misclassifications), supporting a prominent neutral-lipid and phospholipid-remodelling signal and a recurring ether/plasmalogen axis. Findings across both cohorts are consistent with differences in PUFA remodelling, mitochondrial fatty-acid handling, glycerophospholipid metabolism, and amino-acid utilisation/acetylation in LD relative to primary insulin signalling defects, including perturbations in BCAA flux and tryptophan–kynurenine/indole metabolism. The studies in this thesis use complementary strategies to refine cardiometabolic heterogeneity into mechanistically informative endotypes, first identifying common-variant and gene-level contributors to lean-fat partitioning, nominating loci and genes plausibly contributing to healthier tissue distribution and physical resilience across ageing, and second refining metabolomic signatures discriminating lipodystrophic and primary insulin resistance. This advances a framework for moving beyond broad clinical labels towards biologically grounded stratification, while prioritising testable pathways and candidate targets for mechanistic follow-up."]},{"key":"dc:title","label":"Title","values":["Genomic and metabolomic interrogation of body composition and insulin resistance endotypes"]}]}],"canonical_facts":{"dc:contributor.advisor":["Semple, Robert","Cawthorn, William","Loos, Ruth"],"dc:creator":["D'Silva, Sheldon"],"dc:date.accessioned":["2026-06-30T15:55:04Z"],"dc:date.issued":["2026-06-30"],"dc:description.abstract":["Metabolic syndrome and related cardiometabolic disorders are common yet highly heterogeneous, reflecting different ways in which the complex metabolic interplay among insulin-sensitive tissues can go awry. Current clinical classification remains imprecise, however, relying on broad constructs (obesity, type 2 diabetes, BMI thresholds) that do not reliably distinguish underlying mechanisms. One way to inform research into underpinning mechanisms is to anchor it on rare and severe disorders of known cause. In the case of this study this means disorders featuring both severe metabolic derangement and perturbation of body composition. Such rare monogenic disorders reveal causal pathways at the extremes, while population-scale analyses, i.e. genome/exome wide association studies (GWAS/ExWAS), identify common variation shaping risk of more prevalent forms of disease. Healthy skeletal muscle and adipose tissue support physical function and cardiometabolic resilience across the life course. In the general population age-related lean-mass loss and/or disrupted adipose homeostasis are associated with frailty/sarcopenia, insulin resistance (IR), and adverse outcomes. Sentinel monogenic disorders demonstrate that such disproportion between (muscle/adipose) compartments can have a single unifying cause. In the case of increased muscle and reduced adipose tissue this can correlate with either pathological hypofunction i.e. lipodystrophy (LD) or, more rarely, enhanced function and resilience (myostatin deficiency). It was hypothesised that in the general population commoner genetic variants influencing the relative proportion of muscle and fat may also be found, illuminating mechanisms preserving healthy body composition and maintaining metabolic resilience. This thesis first leverages large-scale population genetics, using GWAS/ExWAS to interrogate body-composition phenotypes grounded in observation of paradigmatic monogenic disorders. First, bi-trait analyses of whole body fat mass and fat-free mass normalised by height (FMI and FFMI) were performed using bioelectrical impedance measures from UK Biobank. Variant-level association analyses were conducted for each trait, followed by cross-phenotype tests to identify loci with diverging effects on FFMI and FMI, with locus-level filtering applied to retain signals supported across both traits. This identified 6 variants in men, 5 in women, and 14 in sex-combined analyses. These variants were examined across a panel of cardiometabolic and body-composition phenotypes, and association profiles were compared with rare-variant gene-burden scores in putative mediating genes. Across loci and clusters, PLCE1 and IRS1 showed the most consistent cross-phenotype evidence in the provided analyses, alongside additional supported candidates including PPARG (notably for lipodystrophy-like metabolic correlates) and MTMR11 (for height/anthropometry-heavy signatures), reflecting heterogeneity in the downstream consequences of an FFMI-increasing/FMI-decreasing pattern. A complementary exome-based strategy used gene-based rare-variant burden testing in UK Biobank whole-exome sequencing to analyse association with the composite contrast (FFMI - FMI), and then decomposed significant composite associations into their component FFMI and FMI effects for interpretability. Under the combined functional mask (loss-of-function plus protein-altering missense; alternate allele frequency ≤ 0.001), 25 genes met exome-wide FDR significance for FFMI − FMI while also showing opposite-direction effects on FFMI and FMI. The strongest signals highlighted an extracellular matrix/remodelling theme (including multiple ADAMTS/ADAMTS-like family genes), alongside exemplars with mechanistically informative profiles such as ATP2A1 (lean-mass-aligned with supporting biochemical associations) and GH1 (opposite-direction effects consistent with growth-hormone biology). PLCE1 appeared as a cross-chapter bridge, being supported by both common-variant locus-based analyses and rare-variant gene-based evidence, while emphasising that interpretation and translation should consider known Mendelian disease context for some prioritised genes. The second body of work in this thesis uses metabolomics to assess endophenotypes and disease biomarkers in a population of people with severe IR of known aetiology, mediated by either primary insulin signalling defects or lipodystrophy. Because prognosis and management of these disorders differ sharply, and because diagnosis is often severely delayed, new biomarkers would be valuable. A further motivation was to deepen mechanistic understanding of the differences in these IR subphenotypes to enhance understanding of the common metabolic syndrome. Two independent datasets (Cambridge and NIH) including metabolomic data generated from the same technical platform were analysed. The Cambridge dataset comprised 87 females (LD n = 65; insulin signalling group n = 22). The NIH dataset comprised 59 individuals (LD n = 30; insulin signalling group n = 29). Each cohort was processed separately to avoid cross-cohort artefacts. Metabolite abundances were log-transformed (as provided by the vendor) and robustly scaled, with quality-control checks demonstrating near-identical high-dimensional correlation structure after processing. Highly correlated metabolites were collapsed using medoid reduction. Feature selection used a genetic algorithm to identify metabolite subsets that best separated clinically assigned labels (lipodystrophy vs insulin receptoropathy). Performance was evaluated using cosine-distance-based clustering/classification, summarised by adjusted Rand index (ARI), silhouette score (SS), and misclassification counts, with stability across independent island searches assessed using Jaccard similarity. Across cohorts, the strongest discriminatory structure was carried by lipid and amino-acid domains, while the combined non-lipid/non-amino-acid panel consistently underperformed. In Cambridge, separation was strong for lipids (115-metabolite consensus set; ARI = 0.810; SS = 0.085; 4 misclassifications) and particularly strong for amino acids (138-metabolite consensus set; ARI = 0.951; SS = 0.124; 1 misclassification), whereas the combined panel did not yield meaningful two-group separation (55-metabolite consensus set; ARI close to 0). In NIH, consensus signatures achieved good separation for lipids (144-metabolite consensus set; ARI = 0.804; SS = 0.102; 3 misclassifications) and amino acids (106-metabolite consensus set; ARI = 0.743; SS = 0.177; 4 misclassifications), but only modest separation for the combined panel (66-metabolite consensus set; ARI = 0.263). Complex lipid panels provided some of the cleanest endotype discrimination in both cohorts (Cambridge: 178-metabolite consensus set; ARI = 0.952; SS = 0.231; 1 misclassification; NIH: 171-metabolite consensus set; ARI = 0.804; SS = 0.286; 3 misclassifications), supporting a prominent neutral-lipid and phospholipid-remodelling signal and a recurring ether/plasmalogen axis. Findings across both cohorts are consistent with differences in PUFA remodelling, mitochondrial fatty-acid handling, glycerophospholipid metabolism, and amino-acid utilisation/acetylation in LD relative to primary insulin signalling defects, including perturbations in BCAA flux and tryptophan–kynurenine/indole metabolism. The studies in this thesis use complementary strategies to refine cardiometabolic heterogeneity into mechanistically informative endotypes, first identifying common-variant and gene-level contributors to lean-fat partitioning, nominating loci and genes plausibly contributing to healthier tissue distribution and physical resilience across ageing, and second refining metabolomic signatures discriminating lipodystrophic and primary insulin resistance. This advances a framework for moving beyond broad clinical labels towards biologically grounded stratification, while prioritising testable pathways and candidate targets for mechanistic follow-up."],"dc:identifier.uri":["https://era.ed.ac.uk/handle/1842/44853","https://doi.org/10.7488/era/7367"],"dc:language.iso":["en"],"dc:subject":["Genomics","GWAS","ExWAS","Metabolomics","Insulin Resistance (IR)","IR","Body Composition","Cardiometabolic Disorders"],"dc:title":["Genomic and metabolomic interrogation of body composition and insulin resistance endotypes"],"dc:type":["Thesis"],"dc:type.qualificationlevel":["Doctoral"],"dc:type.qualificationname":["PhD Doctor of Philosophy"]},"updated_at":"2026-07-24T02:14:14Z"}