Ingenuity pathway analysis (IPA)

Prediction-based IPA analysis was used to functionally validate the Meta-dataset in an unbiased way. IPA utilizes advanced literature search techniques from a curated database to predict regulators, mechanistic networks, and sample characteristics based upon the magnitudes and directions of DEG LogFC values. We therefore used IPA to predict characteristics and potential mechanisms of AvN, CvA, and CvN comparison groups to validate that our Meta-dataset indeed represents each sample and potentially their progression. To do this, the list of DEGs between each pairwise comparison determined previously were used for analysis.

First, three disease and function predictions and their corresponding FDR q-values (determined using Fisher’s Exact Test) were used to validate that the signatures corresponded to the correct tissue type and state. Afterwards, the top regulatory networks corresponding to adenoma formation (AvN) and malignant transformation (CvA) were assessed and visualized. Information from the regulatory network including which upstream regulator and downstream pathway was predicted to be active or inhibited was then compared both to the GO analysis (for determining internal consistency) and the literature (for assessing whether results are congruent with what is known).