One of the key questions in bioinformatics and system biology has been the construction of the so-called gene regulatory networks [1], which describe the dependencies that modulate the expression of genes in the living cells. In particular, novel RNA and ATAC measurements at the single-cell resolution appear to be suitable for this inference task.

Recently, this problem has been addressed by various techniques in machine learning, such as the one presented in [2].

The goal of this thesis is to familiarise with the gene regulatory network inference problem in general, and the method presented in [2]. Then, the student should evaluate this method and propose potential improvements, either in terms of computation speed, or in terms of the information extracted from the learned model (e.g. the regulation direction). Subsequently, the method should be used to construct and validate a GRN based on the large scRNA dataset of neural cell differentiation presented in [3].

[1] Mercatelli, Daniele, et al. "Gene regulatory network inference resources: A practical overview." Biochimica et Biophysica Acta (BBA)-Gene Regulatory Mechanisms 1863.6 (2020): 194430.

[2] Keyl, Philipp, et al. "Single-cell gene regulatory network prediction by explainable AI." Nucleic Acids Research 51.4 (2023): e20-e20.

[3] Di Bella, Daniela J., et al. "Molecular logic of cellular diversification in the mouse cerebral cortex." Nature 595.7868 (2021): 554-559.