Background: Genes are coordinately regulated by multiple transcription factors (TFs) or other regulators to form regulatory networks to carry out complex and condition-specific biological functions in human cancer.p53 is the most mutated tumor suppressor gene in cancers, which are usually inflammatory with aberrant NF-B activation.However, how NF-B family members and p53 interact to globally regulate genes expression is not yet fully understood.Methods: Using head and neck squamous cell carcinoma (HNSCC) lines as the model system, we developed a novel integrative model based on Regulatory Component Analysis, which combined mRNA expression profile with regulator (TF and microRNA) binding for integrated analyses through matrix decomposition.The method particularly addresses the sparseness property in the matrix, which is an essential statistical parameter in modeling the regulatory components of regulators and their targets.Results: We identified regulatory programs of seven key TFs, NF-B, p53, AP1, CEBPB, EGR1, SP1 and STAT3, including 37 and 39 common target genes in the cell lines with wild type (wt) and mutant (mt) p53 status, respectively.We observed that the majority of p53 targets are also co-regulated by NF-κB in p53 wt or mt subset of HNSCC cells, suggesting that both TFs play a tight concerted regulation on gene programs in the tumor cells.Furthermore, we further constructed regulatory networks of NF-κB, p53 and microRNAs 21 and 34s.Our results unraveled the cross-regulations among NF-κB, p53, and microRNAs, provided an insight into understanding of underlying regulatory mechanisms.Conclusions: Application of the newly developed method not only showed an efficient approach to inferring the regulatory networks in HNSCC datasets, as the inferred networks are biologically more meaningful than those by other methods, but also provides a model-based pipeline to study on regulatory networks in other biological complex systems .