dbPEC: a comprehensive literature-based database for preeclampsia related genes and phenotypes

Early prediction of preeclampsia in pregnancy with cell-free RNA

Liquid biopsies that measure circulating cell-free RNA (cfRNA) offer an opportunity to study the development of pregnancy-related complications in a non-invasive manner and to bridge gaps in clinical care1-4. Here we used 404 blood samples from 199 pregnant mothers to identify and validate cfRNA transcriptomic changes that are associated with preeclampsia, a multi-organ syndrome that is the second largest cause of maternal death globally5. We find that changes in cfRNA gene expression between normotensive and preeclamptic mothers are marked and stable early in gestation, well before the onset of symptoms. These changes are enriched for genes specific to neuromuscular, endothelial and immune cell types and tissues that reflect key aspects of preeclampsia physiology6-9, suggest new hypotheses for disease progression and correlate with maternal organ health. This enabled the identification and independent validation of a panel of 18 genes that when measured between 5 and 16 weeks of gestation can form the basis of a liquid biopsy test that would identify mothers at risk of preeclampsia long before clinical symptoms manifest themselves. Tests based on these observations could help predict and manage who is at risk for preeclampsia-an important objective for obstetric care10,11.

OUP accepted manuscript

Preeclampsia is a pregnancy-specific disease that can have serious effects on the health of both mothers and their offspring. Predicting which women will develop preeclampsia in early pregnancy with high accuracy will allow for improved management. The clinical symptoms of preeclampsia are well recognized, however, the precise molecular mechanisms leading to the disorder are poorly understood. This is compounded by the heterogeneous nature of preeclampsia onset, timing and severity. Indeed a multitude of poorly defined causes including genetic components implicates etiologic factors, such as immune maladaptation, placental ischemia and increased oxidative stress. Large datasets generated by microarray and next-generation sequencing have enabled the comprehensive study of preeclampsia at the molecular level. However, computational approaches to simultaneously analyze the preeclampsia transcriptomic and network data and identify clinically relevant information are currently limited. In this paper, we proposed a control theory method to identify potential preeclampsia-associated genes based on both transcriptomic and network data. First, we built a preeclampsia gene regulatory network and analyzed its controllability. We then defined two types of critical preeclampsia-associated genes that play important roles in the constructed preeclampsia-specific network. Benchmarking against differential expression, betweenness centrality and hub analysis we demonstrated that the proposed method may offer novel insights compared with other standard approaches. Next, we investigated subtype specific genes for early and late onset preeclampsia. This control theory approach could contribute to a further understanding of the molecular mechanisms contributing to preeclampsia.

Gene-Gene and Gene-Environment Interactions

Identifying gene-gene and gene-environment interactions may help us to better describe the genetic architecture for complex traits. While advances have been made in identifying genetic variants associated with complex traits through more dense panels of genetic variants and larger sample sizes, genome-wide interaction analyses are still limited in power to detect interactions with small effect sizes, rare frequencies, and higher order interactions. This chapter outlines methods for detecting both gene-gene and gene-environment interactions both through explicit tests for interactions (i.e., ones in which the interaction is tested directly) and non-explicit tests (i.e., ones in which an interaction is allowed for in the test, but does not test for the interaction directly) as well as approaches for increasing power by reducing the search space. Issues relating to multiple test correction, replication, and the reporting of interaction results in publications.

Machine learning approach to literature mining for the genetics of complex diseases

To generate a parsimonious gene set for understanding the mechanisms underlying complex diseases, we reasoned it was necessary to combine the curation of public literature, review of experimental databases and interpolation of pathway-associated genes. Using this strategy, we previously built the following two databases for reproductive disorders: The Database for Preterm Birth (dbPTB) and The Database for Preeclampsia (dbPEC). The completeness and accuracy of these databases is essential for supporting our understanding of these complex conditions. Given the exponential increase in biomedical literature, it is becoming increasingly difficult to manually maintain these databases. Using our curated databases as reference data sets, we implemented a machine learning-based approach to optimize article selection for manual curation. We used logistic regression, random forests and neural networks as our machine learning algorithms to classify articles. We examined features derived from abstract text, annotations and metadata that we hypothesized would best classify articles with genetically relevant content associated to the disorder of interest. Combinations of these features were used build the classifiers and the performance of these feature sets were compared to a standard 'Bag-of-Words'. Several combinations of these genetic based feature sets outperformed 'Bag-of-Words' at a threshold such that 95% of the curated gene set obtained from the original manual curation of all articles were extracted from the articles classified by machine learning as 'considered'. The performance was superior in terms of the reduction of required manual curation and two measures of the harmonic mean of precision and recall. The reduction in workload ranged from 0.814 to 0.846 for the dbPTB and 0.301 to 0.371 for the dbPEC. Additionally, a database of metadata and annotations is generated which allows for rapid query of individual features. Our results demonstrate that machine learning algorithms can identify articles with relevant data for databases of genes associated with complex diseases.