Chronic female reproductive disorders such as endometriosis, fibroids, and ovarian cysts can lead individuals to suffer from severe pelvic, abdominal, and/or back pain; heavy menstrual cycles; severe bloating; and reduced fertility. While some individuals with these disorders present with similar symptoms, others can appear asymptomatic until they experience infertility, or the disorder is identified during a medical procedure like tubal ligation. Identifying asymptomatic individuals or differentially diagnosing reproductive disorders for individuals with similar symptoms without pursuing expensive and invasive gold standard diagnostic imaging is difficult. Misdiagnosis and delays in diagnosis are common for reproductive disorders, fueling prolonged symptoms and delays in appropriate therapeutic intervention. Environmental exposures and nutrition have been hypothesized as modifiable risk factors for female reproductive disorders. As such, understanding these risk factors is essential for clinicians to provide recommendations for reproductive disorder prevention and diagnosis. Survey tools are commonly used to collect environmental and nutrition exposure information, but their limited computability makes it challenging to integrate these data with other biomedical data such as phenotypes or genotypes to evaluate reproductive disorder risk. The general objective of this research is to establish new methods for development of semantically computable survey data and their integration with other data modalities and subsequent machine learning for evaluating health risks using biomedical ontologies. The specific goal of this application is to apply these techniques to evaluate clinical data and exposures impacting female reproductive disorders and differentially classify endometriosis, fibroid, and ovarian cyst risk using semantically encoded versus unencoded data. This objective is in alignment with the NICHD Strategic Plan Research Theme of ‘promoting gynecological, andrological, and reproductive health’ through the utilization of ‘integrated genetic and phenotypic exposure data to understand the underlying mechanisms of conditions such as endometriosis, fibroids’. Specific Aim 1 will computationally model female reproductive disorder phenotypes, genotypes, and exposures, focusing on endometriosis, fibroids, and ovarian cysts. Specific Aim 2 will ontologically encode environmental exposures and evaluate reproductive disorder risk, supporting future reproductive research and clinical care improvements. The training plan outlined in this proposal will allow the Candidate to obtain essential training to: 1) Apply novel inference modeling methods for evaluating environmental exposure and nutrition factors influencing reproductive health, 2) Expand skills in data science techniques that can be applied to nutrition and other types of data integration in biomedical sciences, 3) Develop further expertise in nutrition and metabolism and their impact on human reproductive health, 4) Learn methods for clinical and translational requirements analysis and validation approaches, and 5) Develop leadership and career skills required to become a successful independent research scientist.
