New model to help physicians tailor the protocols used for ovarian stimulation

Personalized prediction of the secondary oocyte number after ovarian stimulation.

Share

Infertility is a condition that leads to the failure of natural conception. It affects more than 186 million people worldwide. Because in vitro fertilization (IVF) is an effective infertility treatment, optimizing the steps in the process is essential to assist best those trying to conceive.

The IVF process begins with ovarian stimulation, during which the woman takes ovary-stimulating hormones to produce a certain number of viable, fertilization-ready egg cells, but predicting the number of egg cells collected after such stimulation is difficult. These predictions are usually based on the patient’s clinical parameters and depend on the physician’s experience, making them highly subjective.

A new study- published in PLoS Computational Biology- has developed a model that simultaneously takes advantage of the patient’s genetic and clinical characteristics to predict the stimulation outcome. The model was based on the gradient boosting machine (GBM) technique implemented by the LightGBM framework.

Data for the study was collected between November 2014 and February 2021 at INVICTA Fertility Clinics. The study population consisted of 6,043 women (9,090 IVF processes) diagnosed with infertility undergoing controlled ovarian stimulation with menotropin, follitropin delta, or follitropin alfa.

The study population was divided into two groups according to the availability of genetic data: Group 1, without genetic data, included 5,779 patients and 8,574 IVF processes; Group 2, with genetic data, included 264 patients and 516 IVF processes. All women were between 18 and 46 years old, had regular 26–32-day menstrual cycles, were undergoing their first or second IVF cycle, and exhibited no signs of androgenicity, endometriosis, or any chronic diseases.

Scientists isolated Genomic DNA from whole blood or urogenital swabs. They then used next-generation sequencing to identify sequence variants.

According to the study, the number of MII oocytes in the previous pick-up and the number of cumulus-denuded oocytes in the previous pick-up are also strong predictors of the ovarian stimulation outcome. Combining these features in a single model ensures its high accuracy.

Although studies have suggested that PCOS also affects the number of retrieved oocytes, our study did not identify this condition as relevant for prediction. This discrepancy may be attributed to the low percentage of patients diagnosed with PCOS in our study population.

Single nucleotide polymorphisms have been identified in genes with key roles in oogenesis, folliculogenesis, and female reproduction, such as in estrogen receptor 2 (ESR2), follicle-stimulating hormone receptor (FSHR), FSH β-chain (FSHB), luteinizing hormone β-chain (LHB), LH/choriogonadotropin receptor (LHCGR), growth differentiation factor-9 (GDF9), anti-Mu¨llerian hormone (AMH), and AMH type II receptor (AMHR2) genes. However, the usefulness of sequence variants in clinical practice is limited due to inconclusive results.

The study identifies variants in genes such as ESR1, ESR2, FSHB, FSHR, GDF9, LHCGR, and PRLR correlating with the number of MII oocytes retrieved after ovarian stimulation. Adding these variants to the newly developed model improved its predictive potential. Significant variants were mainly found in genes encoding hormone receptors, for which the association with response to gonadotropins or oocyte maturation has already been confirmed in animal models.

Study authors noted, “Rather than creating a model based on single variants, our approach focused on combinations of variants that were implemented in the model as genetic features and improved the prediction metrics compared to previous studies.”

“Unlike the singly considered sequence variants, the genetic features showed a high occurrence in the population—haplotypes IV22-2 and IV41-8 were found in 58% and 65% of patients, respectively, and an average of 2.6 variants were found within the IV8-6 feature. As a result, many patients could benefit from applying our clinical genetic model.”

“The strength of our study lies in the number of methods used to select sequence variants associated with the number of MII oocytes and the use of multi-variant genetic features instead of single variants for modeling, which increased the predictive potential of genetic data. In contrast to others, our approach was also strengthened by including data on retrospective stimulations.”

“When applied to clinical practice, we believe our model will improve personalized counseling and facilitate decision-making regarding the setting of gonadotropin doses to improve the safety and efficiency of stimulation protocols for IVF.”

Journal Reference:

  1. Zieliński K, Pukszta S, Mickiewicz M, Kotlarz M, Wygocki P, Zieleń M, et al. (2023). Personalized prediction of the secondary oocyte number after ovarian stimulation: A machine learning model based on clinical and genetic data. PLoS Comput Biol 19(4): e1011020. DOI: 10.1371/journal.pcbi.1011020
- Advertisement -

Latest Updates

Trending