Hematological Indices and Genetic Variants of Premature Ovarian Insufficiency: Machine Learning Approaches.

Background: Premature Ovarian Insufficiency (POI) is associated with infertility. Little is known about the potential circulating biomarkers that could be used to predict POI. We have investigated the possible association between white and red blood cells, platelet indices, and eight established single nucleotide polymorphisms (SNPs) associated with POI risk.

Methods: 117 women with premature menopause (PM) and 183 healthy women without a history of menopause before age 40 were recruited for this study. The tetra-primer amplification refractory mutation system-polymerase chain reaction (Tetra ARMS PCR) and allele-specific oligonucleotides- polymerase chain reaction (ASO-PCR) were carried out for genotyping for eight SNPs reported to be associated with POI. Decision tree analysis was applied to test the diagnostic value of hematological parameters to identify the risk of POI.

Results: Women with POI had lower neutrophil (NEUT) and white blood cell (WBC), whereas red blood cell (RBC), hemoglobin (HGB), hematocrit (HCT), mean corpuscular volume (MCV), and mean cell hemoglobin (MCH) were higher. Platelet (PLT) count was also lower in affected women. Our data also indicated that HGB and HCT count were significantly associated with rs16991615 and rs244715. Mean Platelet volume (MPV) and platelet distribution width (PDW) were associated with rs244715, rs1046089, rs4806660, and rs2303369. The rs16991615 was also associated with RBC count, and rs451417 was associated with NEUTs. The decision tree (DT) model reveals that women with the NEUT count at a cut-off value of less than 2.8 and HCT equal to or more than 38.7% could be identified as high-risk cases for POI. Overall, we found the DT approach had a sensitivity = 85%, specificity = 72%, and accuracy = 74%.

Conclusion: The genetic variants involved in POI are associated with changes in reproductive hormone levels and with changes in hematological indices.