4.3. Differential APA Events Detection and Downstream Analysis

The PDUI scores of all samples within the same group were averaged and named as Mean_PDUI_GroupA/B/C. The difference in mean 3′-UTR change per transcript between two groups was quantified by ΔPDUI (ΔPDUI = Mean_PDUI_GroupA − Mean_PDUI_GroupB). A permutation test with looser conditions was used to control the false discovery rate (FDR < 0.05). The differential APA events were defined according to two criterions: ① |ΔPDUI| ≥ 0.1; ② p value < 0.05.

The cluster analysis of the PDUI profiles was conducted using the ComplexHeatmap package [42,43], and the functional analysis of differential APA events was achieved by using Metascape v3.5 [44].

A total of 98 APA factors screened out via literature retrieval [10,45] were subjected to a Spearman correlation analysis with differential APA events, which aimed at detecting possible APAfactors. Additional interactions among them were extracted from the STRING database v.11.5 [46], and the network was visualized using Gephi v.0.10.1 [47]. A network topology analysis was conducted using 12 algorithms (including Betweenness, BottleNeck, Closeness, ClusteringCoefficient, Degree, DMNC, EcCentricity, EPC, MCC, MNC, Radiality, and Stress) via cytoHubba [48] in Cytoscape v.3.9.1 [49].