INFANT Research Centre - Doctoral Theses
Permanent URI for this collection
Browse
Recent Submissions
Item PiRAMiD: predicting early onset autism through maternal immune activation and proteomic discovery(University College Cork, 2023) Carter, Michael; Murray, Deirdre M.; Gibson, Louise; O'Keeffe, Gerard W.; English, Jane; National Children's Research CentreAutism spectrum disorder (ASD) is a heterogeneous developmental disorder arising early in life. ASD is composed of a wide variety of clinical characteristics, neuropsychological impairments and complex phenotypes. The classical triad of ASD symptoms includes disrupted social function, atypical verbal and non-verbal communication skills, and restricted interests with repetitive behaviours. These core symptoms often coexist with other psychiatric and neurological comorbidities. Attention Deficit Hyperactivity Disorder (ADHD), epilepsy, migraine, and anxiety are much commoner in children with ASD. Children and adults with ASD often encounter difficulties with emotional and behavioural problems (EBPs) such as emotional reactivity, aggression, and depression. Up to 50% of those affected can have intellectual disability (ID) and limited verbal communication. Social, emotional and behavioural deficits in children with ASD are also important modifiers of outcome and are linked to elevated stress, mental and physical health problems, and lower overall family and caregiver well-being. We know that early intervention can be effective, and may be parent or therapist delivered. Pharmacological treatment of ASD can be successful insofar as it is useful for symptomatic management of some ASD comorbidities such as ADHD, and depression. Although genetic susceptibilities are increasingly recognised, the mechanism of disease development in ASD remains unknown. We are aware of both common and rare genetic risk factors with more than four hundred diverse high confidence genes now linked to ASD (https://www.sfari.org/resource/sfari-gene/). Singly, these genetic factors each convey only a modest increase in ASD risk (~1%), however collectively they can contribute to a far greater risk. Both de novo and inherited genetic defects are recognised but ASD risk in progeny does not follow a clear pattern of inheritance. Estimates of heritability of ASD in twin pairs vary widely between 50 – 90%. The apparent male preponderance in ASD persists with a clear bias towards males. Rates of ASD among males exceed those of females by three or fourfold hinting at a possible sex differential genetic foundation. Up to 20% of individuals with ASD may possess copy number variants (CNV) and de novo loss of function single nucleotide variants (SNV) that are individually rare but in combination, increase an individual’s overall ASD risk. While newer methods of genetic analysis (such as whole genome sequencing) are uncovering new candidate genes with regularity, the heterogeneity of the clinical and phenotypic groups within ASD strongly suggest that in those with a genetic predisposition, environmental factors may act in concert to bring about a multisystem dysfunction leading to ASD. Despite recent advances in gene analysis, we are yet to discover a single gene determinant that can account for more than a small percent of ASD cases. The current ASD literature suggests that mutations occurring in genes involved in synapse formation, cell adhesion molecule production (Cadherin), scaffolding proteins (SHANK proteins), ion channels (sodium, calcium, and potassium channels), and signaling molecules can disrupt regulatory or coding regions and affect synapse formation, plasticity and synaptic transmission. All this suggests that we cannot explain many cases of ASD by genetic factors alone, or at least we cannot explain them using our current understanding of ASD genetics or our current techniques of genetic analysis. The flawed picture of ASD genetics has led some to investigate the role of environmental exposures in the aetiology of ASD. Researchers have identified many environmental risks in ASD. Advanced parental age, foetal environmental exposures, perinatal and obstetric events, maternal medication use, smoking and alcohol use, psychosocial hardship, nutrition and toxic exposures have all been implicated as risks in the pathogenesis of ASD. While authors attribute between 17 - 41% of ASD risk to non-genetic or environmental exposures, the exact balance between genetic and environmental determinants and their roles in aetiology remains disputed. Multiple mechanisms have been proposed through which each of these exposures may exert an influence on genetic and epigenetic risk in ASD , but there are only a handful that are likely to effect abnormal neurodevelopment. Animal models of inflammation and maternal immune activation are particularly well characterised, and have successfully modelled ASD type behaviours and social difficulties in mice, rats and non-human primates. Maternal immune activation (MIA) is defined as an increase in measured levels of inflammatory markers in mothers during pregnancy. Through this process, a cytokine cascade transmits to the foetus, resulting in adverse neurodevelopmental phenotypes and even remodelling of the immature foetal brain. Many studies have profiled cytokine, chemokine, immune cell and inflammatory signatures in ASD affected individuals. Only a much smaller number have characterised cytokine profiles in expectant mothers who progressed to birth children who develop ASD. The few previous studies, which have examined gestational serum, have indicated mid-gestational upregulation in specific pro-inflammatory cytokines or indeed down-regulation in anti-inflammatory cytokines. Metabolomic analysis refers to the systematic identification and quantitation of all metabolites in a given biological sample. This collection of metabolites, known as the metabolome, is thought to directly reflect the biochemical activity of the studied system at a specific point in time. The metabolome has become an area of interest, as some inborn errors of metabolism (IEM) are clearly linked to ASD phenotypes. Phenylketonuria (PKU) and Smith-Lemli-Opitz syndrome (SLOS) are disorders of amino acid and cholesterol metabolism respectively. Untreated PKU is associated with strongly autistic phenotypes, while SLOS is phenotypically heterogeneous, but autism remains a common feature in these children. Similarly, proteomics is defined as the study of the complete protein profile in a given tissue, cell or biological sample. Proteomic studies of human sera have so far noted altered levels of proteins involved in inflammation or immune system regulation, including acute phase reactants and interleukins. Abnormalities of the complement system have also been found in ASD and other psychopathologies such as schizophrenia. Recent works demonstrate that the complement pathway can affect synaptic remodelling and has roles in neurodevelopmental processes. The initial focus of ASD research on genomics has largely failed to result in the much-hoped-for silver bullet of ASD aetiology, i.e. a common genetic cause. Instead, the genetic landscape has proven to be exceedingly complex and interdependent on a multitude of factors, including environmental exposures and other modifiers of genetic risk. Research examining the aetiology of ASD has shifted focus from genetics to a multimodal approach. In recent years, funding has become available for a far wider variety of ASD aligned research topics, beyond those with a focus on genetics. Opportunities now exist to adopt a multifaceted approach to ASD aetiology, shifting the focus from a narrow genetic base, to a broader multimodal approach to examine other potential mechanistic players. While this adds further complexity to what is already a complicated picture, the strived for parsimonious answer is simply never likely to materialise. Newer fields and modalities such as proteomics, metabolomics and machine learning will help to further refine and untangle the complicated web of ASD, and this variety of granular detail is what is likely to result in a practicable biomarker or effective therapy in the future. In this thesis using a multimodal approach (ELISA, metabolome and proteome analysis) we aim to explore further the role of MIA and alterations of the proteome and metabolome in the pathophysiology of ASD. We hope that our findings may ultimately help to identify a potential gestational biomarker of ASD, which will improve access to early diagnosis and treatment. We also aim to characterise co-morbid emotional and behavioural problems, which arise early in children with ASD and are pervasive throughout all spheres of life. Early recognition and intervention with these co-morbidities can improve treatment outcomes, patient, and family wellbeing.Item A trilogy of stressors in the neonatal intensive care unit: towards therapy for preterm adversity(University College Cork, 2023) Dias Casacao, Maria Luis; McDonald, Fiona; Carolyn Sifton Foundation; Science Foundation IrelandPremature infants are born with immature breathing network and an innate immune system that responds differently to that of infants born full-term. All infants born less than 28 weeks of gestation develop apnoea of prematurity-related symptoms, which decreases to about 20% of infants born at 34 weeks of gestation and to less than 10% of those infants born beyond 34 weeks of gestation. Premature infants are not only at risk of developing breathing disorders but also at increased risk of infection due to early life vulnerability (the earlier a baby is born, the more likely is to have health problems due to immature organ development). Gram-positive (GP) infections are the most common type of late onset infection in preterm infants. Activation of specific toll-like receptors (TLRs) is reported to modulate cardiorespiratory responses during infection and may play a key role in driving homeostatic instability observed during sepsis. Firstly, we sought to characterise the expression of TLRs in the brainstem, adrenal gland and in the diaphragm respiratory muscle in naïve rats during our developmental period of interest (postnatal day (PND) 3 and PND13). These studies demonstrated mRNA expression of these receptors at PND3. TLR expression fluctuated in early life depending on the subtype and tissue examined with a relative decrease in some of the mRNA expression at PND13; TLR1, 2, 4, 6, 9 and NOD2 in sternohyoid muscle, TLR1, 2, 4 and 6 in diaphragm muscle, TLR2 in adrenal glands and NOD2 in brainstem and spinal cord, but a relative increase in mRNA expression of CX3CR1 in brain and brainstem, TLR2 and 9, IL1R1, in brain and TLR2 and TLR9 in spinal cord. Sex differences were revealed in mRNA expression of TLR9 in brain and NOD2 and IL1R1 in brainstem with upregulation of expression in males. These results relating TLRs and postnatal development suggest a developmental regulation of the immune system. Given these results, we reasoned that oxygen dysregulation coupled with GP bacterial immune stimulation would modulate redox sensitive genes and TLR expression that could alter hormonal expression and impinge on respiratory function in a sex-specific manner. To test this, we developed a novel neonatal rat model in which male and female neonatal rats were exposed to intermittent hypoxia, normoxia and hyperoxia from PND3 for 10 days, followed by combined administration of GP bacterial proteins lipoteichoic acid (LTA) and peptidoglycan (PGN). This model sought to mimic physiological challenges encountered by infants born preterm. Hypoxia challenges during the intermittent hypoxia and hyperoxia (IHH) protocol, induced a significant peripheral oxygen desaturation in treated animals. LTA and PGN (3mg and 5mg, respectively) evoked a significant immune response in PND13 rat pups when measured 3 hours post administration. Serum cytokine analysis revealed LTA&PGN triggered an increase in CCL2, IL-1α, IL-1ß, IL-5, G-CSF, IL-13, CCL3, Gro/KC, CXCL10, CX3CL1, CXCL2, IL-10, IFNy, leptin, VEGF, IL-17A and TNF-α serum concentration compared to vehicle. Interestingly, IL-1ß, Gro/KC, IL-10 and leptin expression were upregulated with combined IHH and LTA&PGN exposure. Respiratory function demonstrated an overall decrease in breathing frequency that was mainly impacted by LTA&PGN administration due to an increase in expiratory time. A decrease in minute ventilation was reported with LTA&PGN however, regarding the metabolic function, the ventilatory equivalent for carbon dioxide was similar across the groups, which is consistent with normal pH levels obtained. Additionally, a mild response to the GP challenge in the late periods of hypoxia were associated with decreased number of gasps with IHH. We analysed mRNA expression of TLRs and redox modulated genes using real-time polymerase chain reaction (RT-PCR) in brainstem, diaphragm muscle and adrenal glands. Brainstem gene expression was similar across groups. In adrenal glands tissue, there was an overall upregulation of TLRs mRNA expression with IHH exposure, except for TLR6. Moreover, TLR2 mRNA expression was upregulated with IHH compared to Sham groups, in males compared to females and in LTA&PGN compared to vehicle. An inverse trend from that of adrenal glands was reported in diaphragm muscle. We also analysed redox modulated proteins in serum using bioassays to detect 8-OHdG, 8-iso-PGF2α, AOPP and SOD in plasma. No differences were observed with IHH and LTA&PGN. However, sex differences were found in 8-iso-PGF2α and AOPP redox proteins, with upregulated expression in males compared to females. Finally, we sought to characterise the postnatal development in our animal model using a battery of developmental and motor assessments. These studies demonstrated a delay in pinnae detachment in IHH pups, a decrease in time to righting with IHH and an increase in motor and locomotion faults in IHH females. Tactile stimulation was decrease with IHH suggesting a delay in brainstem reflexes. IHH treated males presented with an increased expression of stress and anxiety-related behaviours illustrated by increased time spent in the corners of the open-field test with LTA&PGN and distance travelled with IHH, and decreased time spent in open arms in the elevated plus maze experiment in males compared to females. Altogether, these results suggest that early life stress can profoundly impact the expression of TLRs and redox genes in adrenal glands, impact the expression of cytokines such as leptin and alter development, motor coordination, stress levels in this novel neonatal rat model of early life stress. In conclusion, these studies specifically target the gaps in knowledge of the pathophysiological alterations experienced in prematurely born babies who present with impaired breathing function and contract GP infection as mimicked in this novel animal model. The results presented disclose novel insights on the physiology and hormonal alterations that these babies could face in similar conditions, with potential to positively contribute to the field of study by enlightening future targets of research. Future directions relying on the spatial characterisation of TLRs and leptin receptors and possible long-term influences on behavioural performance would also be helpful to better characterise the model.Item Axis of placental ageing in adverse pregnancy outcomes(University College Cork, 2023) Manna, Samprikta; McCarthy, Fergus; McCarthy, Cathal; European Chiropractors' UnionBackground: Pre-eclampsia (PE), an adverse pregnancy outcome affects 2-5% pregnancies worldwide and significantly adversely impacts both maternal and fetal outcomes. Intrauterine growth restriction (IUGR) is defined as the inability of the fetus to reach normal growth potential within the uterus as a result of various genetic, environmental, or placental factors. Premature ageing of the placenta in pregnancy outcomes such as PE and IUGR is associated with the persistent presence of oxidative stress and placental insufficiency reducing its functional capacity. Placental proteomics has been instrumental in improving our understanding of molecular mechanisms involved in the pathophysiology of placental insufficiency as well as identifying biomarkers to predict and diagnose pregnancy outcomes. In this study, we investigated cellular senescence phenotypes of PE and IUGR pregnancies by simultaneously measuring several biomarkers of senescence, as well as the proteomic signature of the placenta in healthy and adverse pregnancy outcomes PE and IUGR. Method: Maternal plasma and placental samples were collected at term (>37 weeks) and preterm (<37 weeks) gestation from nulliparous women undergoing prelabour elective Caesarean section with PE without intrauterine growth restriction (PE; n=5), PE associated with intrauterine growth restriction (n=8), intrauterine growth restriction (IUGR <10th centile) (n=6) and age-matched controls (n=20) from Cork University Maternity Hospital, Cork, Ireland. To assess cellular senescence absolute telomere length (aTL) and senescence associated genes in the placentas was performed by RTqPCR. Cyclin-dependent kinase inhibitors (p21 and p16) expression were determined by Western blotting. Senescence Associated Secretory Phenotype (SASP) were evaluated in maternal plasma by multiplex ELISA assay. Proteomic analysis of placental samples dissected into 3 sub-anatomical regions (maternal, middle, fetal) taken from 3 nulliparous healthy placentas was performed by mass-spectrometry and pathway analysis was conducted. Based on the differentially expressed proteins (DEPs), a placenta specific disease map using NaviCenta focusing on functional analysis to include the placenta specific context for healthy (n=4) compared to PE affected (n=4) and IUGR affected (n=4) placentas. Results: Placental expression of senescence associated genes CHEK1, PCNA, PTEN, CDKN2A, CCNB-1 was significantly upregulated in PE, while TBX-2, PCNA, ATM and CCNB-1 expression were significantly decreased in IUGR compared to controls. Moreover, placental p16 protein expression was significantly decreased in PE only when compared to controls placentas. We also observed that IL-6 was significantly increased in maternal circulation in PE when compared to controls; while IFN-γ was significantly increased in maternal circulation in women affected with IUGR when compared to controls. Proteomic profiling of healthy placentas divided into three sub-anatomical regions identified 1081, 1086, and 1101 proteins in maternal, middle, and fetal sub-anatomical regions respectively. Depending on sample site location and sub-anatomical regions, 374 differentially expressed proteins (DEP) were identified. When we investigated the proteomic variations between PE and IUGR placentas when compared to controls we observed 314, 391, and 378 proteins in healthy control, PE, and IUGR placenta, respectively. We performed functional analysis by combining ClusterCompare and NaviCenta to analyse a placenta-centric context, and observed regulatory elements predominantly involved in the immune regulation, complement cascade and antioxidant activities in PE and IUGR compared to control placentas. Conclusion: This thesis provides evidence of premature senescence in IUGR, while in PE, evidence of activated cell cycle checkpoint regulators is suggestive of cellular repair and proliferation rather than progression to cellular senescence. The heterogeneity within senescence molecular markers of these phenotypes highlights the complexity and disparity between pathophysiological insults unique to each obstetric complication. Proteomic profiling of sub-anatomical placental regions highlighted the variabilities between regions particularly providing evidence of senescence in these regions. Placental proteomic mapping of healthy placentas compared to adverse pregnancy outcomes PE and IUGR revealed the importance of complement system, inflammatory response, and antioxidant activity in placental function in PE placentas. The identification of novel targets such as transcription factor activity and synergistic miRNAs elements within the core regulatory network, might enlighten future placental research within adverse pregnancy outcomes.Item Machine learning techniques for identifying early-life biomarkers in perinatal & child health(University College Cork, 2022) O'Boyle, Daragh; English, Jane A.; Murray, Deirdre M.; Mooney, Catherine; Lopez, Lorna; Science Foundation Ireland; Health Research BoardArtificial Intelligence (AI), and more specifically machine learning (ML), has been used in the investigation of biomarkers for many clinical conditions, reducing the need for specialist diagnosis, reducing waiting times and increasing access to reliable diagnostics. There are numerous areas yet to benefit from its application, particularly in the fields of perinatal and paediatric research. Two such brain related conditions, which will be the focus of this thesis are Hypoxic Ischemic Encephalopathy (HIE) and Autism Spectrum Disorder (ASD). HIE is a major cause of neurological disability globally and results from a lack of oxygen to the brain during and immediately after birth. The mainstay of treatment is therapeutic hypothermia, which, to be effective it must be applied within 6 hours of birth. This thesis aims to improve the early identification of infants eligible for therapeutic hypothermia using AI with clinical and metabolomic HIE biomarkers. Autism constitutes a group of neurodevelopmental disorders characterized by behavioural and cognitive symptoms. The underlying aetiology behind autism remains unclear and reliable predictive biomarkers are lacking. Intervention from an early age has been shown to reduce symptoms but diagnosis frequently occurs outside the window for effective treatment. Despite proven benefits in patient outcomes with intervention within the first two years of life, diagnosis often doesn’t occur until an individual is 3 or 4 years old, and in many cases much older. There is a pressing need for new methods to identify neonates most at risk to provide adequate treatment which improves long term outcomes. In the first experimental section, chapters 2-5, we have applied ML to first identity the optimum predictive clinical and metabolomic biomarkers for the identification of those with HIE. We initially assessed clinical variables using ML feature ranking and modelling. This study identified markers of a newborn’s condition at birth; Apgar scores, need for resuscitation, first measurements of pH and base deficit, as the most predictive. These models achieved an area under the receiver operator characteristic curve (AUROC) of 0.89 when distinguishing between those with perinatal asphyxia, who do not require treatment, and those with HIE. Furthermore, we then assessed a panel of promising metabolite markers for their predictive capabilities with and without clinical markers. ML identified metabolites alanine and lactate as the most predictive of HIE development and when combined with Apgar scores a measure of a newborn’s condition, at 1 and 5 minutes after birth, achieved a predictive AUROC of 0.96. These studies have successfully identified alanine as a candidate metabolite for cotside HIE risk assessment as well as displaying that ML models can improve on our current ability to identify those in need of therapeutic hypothermia. To validate these findings two studies were undertaken. The first independently compared alanine levels in cord blood of those with HIE and controls. This study has displayed elevated levels of alanine at birth and up to 6 hours after, successfully validating alanine as an early life marker for those with HIE in need of neuroprotective therapy. The second validation study successfully validated our algorithm for the prediction of HIE in a large diverse cohort comprised of infants with a range of differing conditions, compared to a set of HIE cases and controls previously assessed. Here 243 infants were assessed, using our model, to determine risk of HIE and an accuracy of 85% was maintained. This study has successfully validated our model's ability to retain performance when applied to a diverse, real-world cohort. In this section, we have successfully displayed the use of ML for improving HIE diagnostics and validated these findings. Further, larger validation studies are currently underway with the end goal of clinical use for determination of those in need of treatment for HIE. In the second experimental section, chapters 6-8, we aimed to apply ML methods to identify early life biomarkers for autism spectrum disorder. We first conducted a systematic review of all blood-based autism biomarkers. This study successfully catalogued all reported biomarkers and recorded the direction of change of theses markers in those with autism compared to neurotypical controls. This study also applied Genome Wide Association Studies (GWAS) and pathway analysis to test for biological processes which may be implicated at the level of the genome in autism. In chapter 7, ML analysis was applied to metabolomics data from cord blood samples from the Cork BASELINE birth cohort. Discovery and targeted metabolomics were completed on this data. In chapter 8, we applied ML to assessed clinical predictors for autism in the Danish National Birth Cohort, which included data from 500 autism cases and matched controls. We identified markers of maternal health and wellbeing as being important for autism prediction and achieved a prediction accuracy of 0.68 AUROC. Overall, this thesis successfully addressed its’ aims to apply ML methods for the identification of biomarkers and development of prediction models for HIE and autism. We have validated previously identified biomarkers and identified novel clinical and blood-based markers as well as created robust HIE predication models with the ability to improve clinical decision making. Possible future steps this research can follow to further add to the field are outlined within. Overall, this research has added to the growing body of evidence displaying the ability of ML to offer improvements to healthcare and specifically to perinatal and child health. Artificial Intelligence (AI), and more specifically machine learning (ML), has been used in the investigation of biomarkers for many clinical conditions, reducing the need for specialist diagnosis, reducing waiting times and increasing access to reliable diagnostics. There are numerous areas yet to benefit from its application, particularly in the fields of perinatal and paediatric research. Two such brain related conditions, which will be the focus of this thesis are Hypoxic Ischemic Encephalopathy (HIE) and Autism Spectrum Disorder (ASD). HIE is a major cause of neurological disability globally and results from a lack of oxygen to the brain during and immediately after birth. The mainstay of treatment is therapeutic hypothermia, which, to be effective it must be applied within 6 hours of birth. This thesis aims to improve the early identification of infants eligible for therapeutic hypothermia using AI with clinical and metabolomic HIE biomarkers. Autism constitutes a group of neurodevelopmental disorders characterized by behavioural and cognitive symptoms. The underlying aetiology behind autism remains unclear and reliable predictive biomarkers are lacking. Intervention from an early age has been shown to reduce symptoms but diagnosis frequently occurs outside the window for effective treatment. Despite proven benefits in patient outcomes with intervention within the first two years of life, diagnosis often doesn’t occur until an individual is 3 or 4 years old, and in many cases much older. There is a pressing need for new methods to identify neonates most at risk to provide adequate treatment which improves long term outcomes. In the first experimental section, chapters 2-5, we have applied ML to first identity the optimum predictive clinical and metabolomic biomarkers for the identification of those with HIE. We initially assessed clinical variables using ML feature ranking and modelling. This study identified markers of a newborn’s condition at birth; Apgar scores, need for resuscitation, first measurements of pH and base deficit, as the most predictive. These models achieved an area under the receiver operator characteristic curve (AUROC) of 0.89 when distinguishing between those with perinatal asphyxia, who do not require treatment, and those with HIE. Furthermore, we then assessed a panel of promising metabolite markers for their predictive capabilities with and without clinical markers. ML identified metabolites alanine and lactate as the most predictive of HIE development and when combined with Apgar scores a measure of a newborn’s condition, at 1 and 5 minutes after birth, achieved a predictive AUROC of 0.96. These studies have successfully identified alanine as a candidate metabolite for cotside HIE risk assessment as well as displaying that ML models can improve on our current ability to identify those in need of therapeutic hypothermia. To validate these findings two studies were undertaken. The first independently compared alanine levels in cord blood of those with HIE and controls. This study has displayed elevated levels of alanine at birth and up to 6 hours after, successfully validating alanine as an early life marker for those with HIE in need of neuroprotective therapy. The second validation study successfully validated our algorithm for the prediction of HIE in a large diverse cohort comprised of infants with a range of differing conditions, compared to a set of HIE cases and controls previously assessed. Here 243 infants were assessed, using our model, to determine risk of HIE and an accuracy of 85% was maintained. This study has successfully validated our model's ability to retain performance when applied to a diverse, real-world cohort. In this section, we have successfully displayed the use of ML for improving HIE diagnostics and validated these findings. Further, larger validation studies are currently underway with the end goal of clinical use for determination of those in need of treatment for HIE. In the second experimental section, chapters 6-8, we aimed to apply ML methods to identify early life biomarkers for autism spectrum disorder. We first conducted a systematic review of all blood-based autism biomarkers. This study successfully catalogued all reported biomarkers and recorded the direction of change of theses markers in those with autism compared to neurotypical controls. This study also applied Genome Wide Association Studies (GWAS) and pathway analysis to test for biological processes which may be implicated at the level of the genome in autism. In chapter 7, ML analysis was applied to metabolomics data from cord blood samples from the Cork BASELINE birth cohort. Discovery and targeted metabolomics were completed on this data. In chapter 8, we applied ML to assessed clinical predictors for autism in the Danish National Birth Cohort, which included data from 500 autism cases and matched controls. We identified markers of maternal health and wellbeing as being important for autism prediction and achieved a prediction accuracy of 0.68 AUROC. Overall, this thesis successfully addressed its’ aims to apply ML methods for the identification of biomarkers and development of prediction models for HIE and autism. We have validated previously identified biomarkers and identified novel clinical and blood-based markers as well as created robust HIE predication models with the ability to improve clinical decision making. Possible future steps this research can follow to further add to the field are outlined within. Overall, this research has added to the growing body of evidence displaying the ability of ML to offer improvements to healthcare and specifically to perinatal and child health.Item Rethinking stillbirth through behaviour change(University College Cork, 2022) Escañuela Sánchez, Tamara; O'Donoghue, Keelin; Matvienko-Sikar, Karen; Meaney, Sarah; Byrne, Molly; Science Foundation IrelandBackground Worldwide, two million babies are stillborn every year. While the majority of stillbirths occur in low and middle-income countries, stillbirth is still one of the most common adverse pregnancy outcomes in high-income countries. In Ireland, the latest National Perinatal Mortality Clinical Audit report states a stillbirth rate of 4.20 per 1000 births for the year 2020, showing an increase compared to previous years. The belief that reduced stillbirth rates in high-income countries cannot be achieved is refuted by differences in stillbirth rates across different countries. Although not all stillbirths are preventable, there has been a call made in high-income countries to focus on risk factors for stillbirth, in order to reduce stillbirth rates. These risk factors include sociodemographic factors, medical factors, obstetric history-related factors, placental and fetal-related factors as well as behavioural and lifestyle-related factors. Some of these factors are modifiable through medical management or through behaviour change modification. This Thesis focuses on risk factors that have the potential to be modified through maternal behaviour change interventions: substance use (smoking, alcohol, and illicit drug use), high BMI, sleep position, and attendance at antenatal care. Strategies have been successfully implemented internationally to reduce stillbirth rates by designing and implementing care bundles that, amongst other elements, take into consideration the modifiable/behavioural risk factors for stillbirth. However, in Ireland, no such initiatives have been developed, although recommendations have been made that support their development. For behaviour change interventions or public health initiatives to have the best possible success in reducing the rates of stillbirth, they need to be designed with a solid evidence base. Hence, the overall objective of this Thesis was to build the evidence base to enhance the understanding of the modifiable behavioural risk factors for stillbirth and pregnancy. Further, this evidence base is needed to inform the future development of a behaviour change intervention that could be part of a care bundle with the objective of reducing stillbirth rates in Ireland. Methodology To address the Thesis´s aims, both qualitative and quantitative methods were utilised. Applying multiple methods to explore a phenomenon provides flexibility to analyse different aspects of it in the different studies. Initially, a non-systematic review of the literature was conducted to identify the target behavioural risk factors that this project was going to focus on (Chapter 2). A website quantitative content analysis was conducted to assess the availability of information related to stillbirth and behavioural risk factors for stillbirth in Irish and UK websites (Chapter 3). For this study, descriptive and inferential statistics were utilised. Further, three systematic qualitative meta-synthesis were conducted to identify facilitators and barriers to modify identified behavioural risk factors according to the pregnant women’s experience (Chapters 4-6). A meta-ethnographic approach as described by Noblit and Hare was adopted to conduct these qualitative meta-syntheses. Reflexive Thematic Analysis as described by Braun and Clarke, with a constructivist approach, was used to conduct a qualitative semi-structured interview study with postpartum women about their experiences of stillbirth information provision and behaviour change during their antenatal care (Chapter 7). Finally, a systematic review of interventions designed in the context of stillbirth prevention that targeted behavioural risk factors was conducted (Chapter 8). This systematic review had the objective of identifying which behaviour change techniques (BCTs) have been used to date. Results The findings of the literature review (Chapter 2) showed that the modifiable behavioural risk factors with the strongest evidence of associations with stillbirth were substance use, smoking, heavy drinking and illicit drug use, lack of attendance and compliance with antenatal care, weight-related risks, and sleep position. The quantitative content analysis of websites (Chapter 3) revealed that information about stillbirth and behavioural risk factors for stillbirths was scarce on websites directed at the pregnant population, with only one website containing all the information sought. Five main areas of concern were identified across the three meta-synthesis of qualitative research of facilitators and barriers influencing women’s prenatal health behaviours (Chapters 4-6), regardless of the behaviour explored: 1) health literacy, awareness of risks and benefits; 2) insufficient and overwhelming sources of information; 3) lack of opportunities and healthcare professionals attitudes interfering with communication & discussion; 4) social influence of environment, and 5) social judgement, stigmatisation of women and silence around stillbirth. Further, the qualitative study with postpartum women (Chapter 7) revealed that women perceived behaviour change during pregnancy as easy and natural, as they were focused on obtaining the best outcomes for their babies. Although women had high levels of awareness regarding health advice, their awareness about stillbirth was very limited. Women reported a lack of discussion about stillbirth and behavioural risk factors during their antenatal care; however, most women showed a positive disposition towards receiving this information because “knowledge is key”, as long as it is done in a “sensible manner”. The systematic review of interventions designed in the context of stillbirth prevention identified nine relevant interventions. From the BCT coding, it was established that the most common BCT used was “information about health consequences”, followed by “adding objects to the environment” (Chapter 8). Conclusion This research makes a valuable contribution to the understanding of the maternal behaviours associated with an increased risk of stillbirth, and it provides a necessary evidence-base to inform future prevention strategies to reduce rates of stillbirth in Ireland and in similar healthcare settings. This research sought to incorporate women’s voices and use research methods to produce high-quality results that meet the research objectives. The findings from the studies in this Thesis support four overarching topics and highlight issues related to 1) health literacy and sources of information, 2) relationships with healthcare professionals (HCPs), 3) healthcare systems and structural barriers, and 4) interpersonal, social and structural factors. In response to the research findings, several recommendations are made in relation to policy, practice and research which are grounded on women’s experiences during pregnancy. Regarding policy, these recommendations include improving education and information sources for women and HCPs, providing pregnancy-specific supports, utilising community services to support women with behaviour change, and developing a care bundle to tackle the behavioural risk factors for stillbirth. Furthermore, the work practice recommendations made include developing clinical guidelines to support HCPs in providing care to pregnant women, and prioritising health promotion during antenatal care. These priorities might also serve to help funders and researchers to design and conduct policy-relevant research. The key future research areas identified by this Thesis are in relation to the involvement of PPI representatives, the assessment of the quality of the available sources of information and the further exploration of potential facilitators and barriers to modifying pregnant women’s sleeping position from a qualitative perspective. In addition, this Thesis proposes a detailed process to continue building on the work set out in the different studies to develop a pregnancy-specific behaviour change intervention for the modifiable behavioural risk factors for stillbirth in the future.