Mid-pregnancy immune dysregulation and its association with maternal metabolic and genetic factors in those with preterm birth: Insights using artificial intelligence.
AAI Annual Meeting; May 6-10; Portland, OR. 2022
oregon; portland; chiles; genomics
Introduction: Mid-pregnancy immune dysregulation, as indicated by maladaptive cytokine and immune cell signaling, is associated with an increased risk of preterm birth (PTB) (birth at < 37 weeks of completed gestation). While infection and immune-related disease (e.g. asthma, type I diabetes) explain some portion of observed immune dysregulation in pregnancies with PTB, in most instances this dysregulation is unexplained. Investigation of the relationship between immune dysregulation and other molecular factors with known links to PTB (e.g. metabolic and genetic factors) offers the opportunity for revealing novel causal pathways.
Methods: In a sample of pregnant women with PTB (n = 482, <37 weeks) and term birth (n = 322, 39-42 weeks), we explored the relationship between mid-pregnancy immune dysregulation and metabolic and genetic factors using causal artificial intelligence (AI) methods. Analyses were divided into 2 phases wherein I: candidate metabolic and SNP factors that differed in those with and without PTB were identified; And II: the relationships between candidate SNP and metabolic factors with immune dysregulation (as determined by an established network of cytokine and immune cell signals) were examined. In phase I, metabolic and genetic factors were identified from a set of 745 metabolites and 651,000 SNPs with retention based on chi-square analysis (p < .001 after False Discovery Rate (FDR) correction, or unadjusted p < 10-5 for SNPs) and the use of a parallelized grid search strategy that identified the best classifiers independent of p-values. In phase II, the relationship between immune dysregulation, metabolic factors, and SNPs retained in phase I were further explored using Bayesian causal network methods.
Results: Using chi-square and parallelized grid search strategies, 28 metabolic, and 60 SNPs were found to be associated with PTB. Within this set, six metabolites (13-HODE + 9-HODE, 2-hydroxystearate, tryptophan betaine, palmitate (16:0), beta-hydroxyisovalerate, N1-methyladenosine), and five SNPs (2:216734070, 5:5181928, 7:151459336, 8:4481743, 9:132946644) were found to be related to immune dysregulation (significantly associated with five or more markers contained within established immune dysregulation models for PTB) (Figure). A number of cross metabolite, cross SNP, and metabolite by SNP associations were observed within the immune dysregulation network. For example, palmitate (16:0) was positively associated with 13-HODE + 9-HODE, 2-hydroxystearate, and beta-hydroxyisovalerate (Pearson’s Correlation Coefficients (PCCs) 0.16-0.53, p < .05) and negatively associated with N1-methyladenosine and 8:4481743 (PCCs -0.09 and -0.10 respectively, p < .05) (Figure).
Conclusions: Here we demonstrate a novel approach to investigating predictors and drivers of PTB and associated immune dysregulation. Our findings confirm the association of immune, metabolic, and genetic factors with PTB and identify new cross-pathway relationships that may be mechanistically important.
Women & Children
Earle A. Chiles Research Institute
Obstetrics & Gynecology
Jelliffe-Pawlowski, L; Piening, Brian D.; and See full list of authors in comments, "Mid-pregnancy immune dysregulation and its association with maternal metabolic and genetic factors in those with preterm birth: Insights using artificial intelligence." (2022). Articles, Abstracts, and Reports. 6110.