MICA: A multi-omics method to predict gene regulatory networks in early human embryos

Abstract

Recent advances in single-cell-omics have been transformative to characterise cell types in challenging to study biological contexts. It is technically difficult to infer transcription factor-gene regulatory interactions from these datasets, especially in contexts with limited single-cell sample number such as the early human embryo. Here we systematically assessed the application of four different linear or non-linear gene regulatory network (GRN) prediction strategies to single-cell simulated and human embryo transcriptome datasets. We also compared how gene expression normalisation methods impact on regulatory network predictions, finding that TPM (transcripts per million reads) outperformed alternative methods. We identified more reproducible GRN inferences using a non-linear method based on mutual information (MI) applied to single-cell transcriptome datasets refined with chromatin accessibility (CA) (called MICA), compared to alternative network prediction methods tested. MICA captures complex non-monotonic dependencies and feedback loops. Using MICA, we generated the first GRN inferences in early human development. MICA predicted co-localisation of the AP-1 transcription factor subunit proto-oncogene JUND and the TFAP2C transcription factor AP-2{psi} in human preimplantation embryos. These interactions were supported by previous findings in other developmental and stem cell contexts. Overall, our comparative analysis of gene regulatory network prediction methods defines a pipeline that can be applied to single-cell multi-omics datasets in especially challenging contexts to infer interactions between transcription factor expression and target gene regulation.