Kuijjer Lab

New BioRxiv pre-print

April 15, 2022

We have a new BioRχiv pre-print, describing our latest tool PORCUPINE and its application to leiomyosarcoma, an aggressive soft-tissue sarcoma subtype. PORCUPINE, which was developed by Tatiana, analyzes complex, genome-wide, patient-specific regulatory networks directly on the network's edge weights. It does this by combining Principal Component Analysis (PCA) with permutations to identify pathways contributing to regulatory heterogeneity in a patient population. Importantly, instead of identifying discrete subtypes, it can map subtle gradients of heterogeneity that may not be detected by standard subtyping approaches.

    Applying PORCUPINE to leiomyosarcoma data, we identified 37 heterogeneously regulated pathways, including targetable pathways such as E2F signaling. These findings were independent of mutations and could thus be a new way of stratifying patients for personalized medicine. The BioRχiv pre-print can be found here. More information on the pre-print can be found here.
PORCUPINE extracts edge weights connected to genes belonging to a specific biological pathway, in patient-specific network models. It applies PCA, extracts the variance explained by the first principle component. It repeats this approach randomly generated gene sets of the same size as the selected pathway, and estimates the effect size and significance that the selected pathway explains a higher proportion of regulatory heterogeneity than expected by chance. After doing this for all pathways, it corrects p-values for multiple testing and returns those pathways that are significantly heterogeneously regulated in a patient population.