We developed sparseGMM, the next evolution of the previously developed GRN inference method called LINKER , a module network approach in a Bayesian framework, whereby the clustering of target genes and the assignment of regulators are combined in one step, which allows genes to be associated to multiple modules simultaneously. The assignment of a regulator to its modules can be thus calculated with a confidence interval. We show an improved performance in sparsity, compared to previous methods, choosing fewer genes as true regulators, and confirming biological knowledge of the scale-free nature of gene networks. Further, our probabilistic assignment approach is potentially superior for modeling genes with multiple biological functions. Thus, we define the entropy of a gene to be the entropy of the estimated module-assignment probability and show that it can then be used as an indicator of a multifunctional biological role based on joint membership to two or more modules. These multifunctional genes could in turn translate to multifunctional proteins having central roles in the crosstalk between two or more pathways in cancer cells, and, thus become attractive targets for overcoming drug resistance through compensation mechanisms.
We show that high-entropy genes are more common in cancer samples than in healthy tissue, and we associate them to crosstalk between several pathways including TP53, interferon gamma and TNF alpha. Our analysis of high entropy genes exemplifies ways in which major cancer pathways share key multifunctional components.
Evaluating the quality and accuracy of GRNs is not a straightforward task, as there are no clear metrics for evaluation. In addition, GRNs are difficult to interpret and visualize. While synthetic data is typically used to evaluate their goodness of fit to the data, it has been shown that GRNs excelling on synthetic data do not necessarily provide more accurate biological insights. To address this issue, we developed a novel methodology to evaluate the altered regulatory dynamics of the different TFs across the different cell phenotypes. The developed SimiC workflow can generate a heatmap that shows the regulatory dissimilarities for all regulons and all cell clusters. which allows us to uncover shifts in regulatory activity that are associated with different conditions, environments, or developmental states.
When applied to hematological malignancies, we found specific biological processes for each hematological differentiation trajectory that are potentially abrogated in disease with respect to healthy samples. Example of these are the Heme-metabolism and gas transport in the Erythroid-Megakaryocyte differentiation trajectory, or Neutrophil activation and degranulation in the Monocytic-Granulocytic differentiation trajectory. In addition, we highlighted two transcription factors: ZNF350 and ZMAT2. These TFs significantly downregulated all their target genes in Myelodisplastic síndromes, whereas in the healthy condition no significant correlations were recorded.
These results were presented at both national (SEHH 21) and international (ASH 2021) clinical haematology conferences, and the computational model were presented at the International Conference of Intelligence Systems for Molecular Biology (ISMB 2021). The action also resulted in several publications. Finally, since all developed computational models can be applied to a myriad of biomedical problems, all developed methods are openly available in open source repositories to facilitate their use by the wider biomedical research community.
