- Meeting ID: 867 6409 6440
- passcode: 149120
13h00 - 13h50 – Célia Messaoudi (LS2N, Nantes, COMBI Team)
Improving Boolean network inference and robust solutions identification : application to human embryology
ABSTRACT Understanding human embryonic development is essential to improve IVF (In Vitro Fertilization) protocols, as only 25% of them lead to a viable pregnancy. The critical step being the implantation of the embryo in the uterus, we chose to focus on the TE (Trophectoderm), which is the outer cell layer responsible for the attachment of the embryo to the endometrium. Specifically, we study the medium and late stages of the TE using Boolean networks inferred from single cell transcriptomics data. We used the SCIBORG tool to infer boolean network families from scRNA-seq data and prior knowledge. However, some default parameters used in SCIBORG are arbitrary and lead to the inference of very similar network families. In this work, we present two improvements of the SCIBORG pipeline : (i) an updated prior knowledge network (PKN) using Pathway Commons v14, (ii) a gene-specific Gaussian Mixture Model (GMM) binarization instead of the fixed threshold used before. Using an adapted Jaccard index and a stage discrimination step, we compared solutions inferred with both binarization functions as well as already published ones, and the solutions with the best BAC scores were kept. We find that the GMM-based approach yields fewer solutions but they discriminate the two stages well with BAC scores reaching 72%. We also present how to use quality-assessing indexes and scores to identify the best solutions.
Dernière modification le 06/02/2026