Gene expression profiles from 844 multiple myeloma patients were used to create a transcriptomic landscape (tSNE, a) of multiple myeloma featuring 16,738 protein coding genes, where co-expressing gene programs using fuzzy c-means clustering (b) are identified. These unsupervised gene clusters/programs represent an undetermined biology that describes multiple myeloma. Co-expression gene clusters likely involve a common upstream regulators governing the expression of their target genes through cooperative or competitive binding. Several upstream regulators (TFs) are post-translationally modified via kinases, many of them are druggable. This offers a mechanistic pathway to target certain co-expression gene clusters associated with a certain disease pathobiology or therapy resistance via novel therapies. To this end, a gene regulatory network is constructed with cascading layers of target genes (co-expressing gene cluster) their upstream transcription factors informed by publicly available ChIP-seq databases like ENCODE and ChEA, and kinases that phosphorylate the transcription factors using PhosphoPoint and PhosphoSitePlus, as shown in c. The functional relationships between each of the nodes in this network are governed by a biophysical model of nonlinear reaction kinetic equations accounting for transcription (d), translation (e), and post-translational modifications (f). At steady state, we can solve for the gene expression of a target gene (G_i) from gene expression of upstream transcription factors (G_TF) and kinases (G_K) along with reaction rate constants informed by patient specific KEGG pathway enrichment scores, as shown in g. This biophysical model was trained on 50% of the cohort and validated on the other 50% to yield correlations between model predicted and actual gene expressions, where genes with accurate prediction (r>0.5) are shown in red in h. Interestingly, the genes with inaccurate model predictions were significantly described by varying chromatin accessibility from (ATAC data) between patients, thereby implying an epigenetic role in regulating their expression.

​

​​

Owing to the biophysical model describing the relationship between target genes and their upstream regulators and kinases, the model can simulate the effect of kinase inhibition on target genes. Thereby, the model can help elucidate the role of novel targeted therapies in reversing resistance to standard-of-care drugs.​​