Hierarchical representation learning of preeclampsia interactome connecting endometrial maturation, placentation, chorioamnionitis, and HELLP syndrome

BackgroundExisting proposed pathogenesis for preeclampsia (PE) was only applied for early-onset PE (EOPE). Our previous work identified the transcriptome to decipher EOPE and late-onset PE (LOPE), but the pathogenesis models were not validated. We developed and validated the pathogenesis models by hierarchical representation learning of interactomes connecting endometrial maturation, placentation, chorioamnionitis, and hemolysis, elevated liver enzyme, and low platelet (HELLP) syndrome. MethodsWe utilized 19 gene sets from our previous work to infer interactomes to develop (n=177) and validate (n=352) explainable artificial intelligence models for each PE subtype using deep-insight visible neural network and gradient-weighted class activation mapping. ResultsThe hierarchically learned representations identified novel genes for LOPE, similar to endometrial maturation (MRPL34, DYNLL1), chorioamnionitis (ANKRD13A, SLA), and HELLP syndrome (FAM43A). We also identified novel genes for EOPE, similar to endometrial maturation (SNAP23, PPL, LRRC32), placentation (GPT2, UBE2H, NIPAL3, NIN, KIAA0232, MT1F, DKK3, SLC24A3), and HELLP syndrome (SWAP70, GREM2, GPR146, PIP5K1B, EZR). Nonetheless, a gene for each subtype was frequently studied, i.e., IGF1 (chorioamnionitis) and PAPPA2 (placentation), including LOPE and EOPE samples. ConclusionsOur pathogenesis models connected both endometrial maturation and HELLP syndrome with LOPE and EOPE. However, they were differently connected with chorioamnionitis and placentation.