Cancer is rarely the straightforward consequence of an abnormality in a single gene, but rather reflects a complex interplay of many genes, represented as gene modules. Here, we leveraged the recent advances of model-agnostic interpretation approach and developed CGMega, an explainable and graph attention-based deep learning framework to perform cancer gene module dissection. CGMega outperforms current approaches in cancer gene prediction, and it provides a promising approach to integrate multi-omics information. We applied CGMega to breast cancer cell line and acute myeloid leukemia (AML) patients, and we uncovered the high-order gene module formed by ErbB family and tumor factors NRG1, PPM1A and DLG2. We identified 396 candidate AML genes, and observed the enrichment of either known AML genes or candidate AML genes in a single gene module. We also identified patient-specific AML genes and associated gene modules. Together, these results indicate that CGMega can be used to dissect cancer gene modules, and provide high-order mechanistic insights into cancer development and heterogeneity.