Fig. 4
(A) GraphComm can identify differences in interactions among datasets that have the potential to be indicative of change in condition. Using multiple datasets of PC9 lung adenocarcinoma cell lines sequenced at different timepoints before and after treatment with osimertinib including two biological replicates, the fraction of overlapping interactions can be compared. This allows for the interpretation of how GraphComm can (i) identify more unique interactions in two datasets of different condition than two biological replicates and (ii) identify more common interactions in two biological replicates than two datasets of different condition. (B) Between one PC9 dataset at day 0 (pre-treatment) and one PC9 dataset at day 7 (post-treatment), GraphComm is able to prioritise unique top-ranked interactions, with an overlap of 56%. (C) In comparison to overlap between pre and post treatment datasets, GraphComm achieves a greater overlap between two day 7 (post treatment) biological replicates, with an overlap of 72%. (D) In comparison to top-ranked interactions of 100 randomised iterations, GraphComm achieves a larger overlap difference of biological replicates vs pre and post treatment datasets using the original directed graph and ground truth. (E) In comparison to top-ranked interactions of 100 randomised iterations, GraphComm achieves significantly better overlap between two biological replicates using the original directed graph and ground truth. Created in BioRender. So, E. (2025) https://BioRender.com/f51h099.
