Genes regulate one another to control cell behavior, and disruptions in these pathways are a primary cause of diseases like cancer. Mapping these interactions into a graph structure called a Gene Regulatory Network (GRN) remains an ongoing challenge due to noisy biological data. However, the increasing availability of gene expression and single-cell data now enables the exploration of Deep Neural Networks (DNNs) in this field. The Biology-informed Robust AI (BRAI) project creates artificial GRNs based on known biological properties, such as sparsity and degree distributions. Simulated "labeled" gene expression data can be generated from these artificial GRNs and used for semi-supervised learning of DNNs. Training models on both simulated and real data aims to better infer these networks and uncover new insights about cancer.