Multiple myeloma (MM) is a clonal plasma cell malignancy with a heterogeneous genetic background. Extensive gene expression profile analysis have provided interesting insight into the disease biology and its correlation with clinical outcome; however, we have begun to realize the significant limitations of expression profile data alone. Therefore there is a growing understanding that additional genomic correlates need to be incorporated to develop an integrated oncogenomic analysis. We have here developed and defined multi-gene transcriptional and post-transcriptional feed-forward loop (FFL) These conceptual FFLs consist of a master TF which regulates a miR and together with it controls a set of specific common gene/s. These recurrent and important network motifs form functional nodes in the larger regulatory network, and are considered linchpins of disease causing genomic alterations in cancer and MM in particular. We have developed a comprehensive novel integrative analysis method, dChip-GemiNI (Gene and miRNA Network-based Integration), which combines gene and miR expression profiles, and also incorporates regulatory network structure in the form of computationally identified TF–miRNA FFLs. The dChip-GemiNI method statistically ranks computationally predicted FFLs by their explanatory power to account for differential gene and miRNA expression. We have next applied dChip-GemiNi to a training dataset of 60 MM patients and 5 normal plasma cells (NPCs) with both gene expression (GE) and miR profiles (dataset GSE16558) in order to identify FFLs containing TF-miR-gene networks with loss of negative feedback regulation in MM, supporting the uncontrolled growth, anti-apoptosis and/or other oncogenic effects. We have identified 20 FFLs significantly aberrant between NPC and MM cells. Prominent FFLs involve known MM dysregulated TFs such as MYC, TP53 and Sp1. In addition, we have utilized 3 available myeloma datasets with both miR and GE profiles (GSE16558, GSE17306, GSE17498), and classified MM samples into hyperdiploid MM (HMM) and non-hyperdiploid MM (NHMM) subtype groups by GE profiles at an accuracy >85%. These two groups have different survival outcomes (p-value < 0.01). We have identified 55 FFLs altered between these two MM subtypes. In particular we have observed that the FFL involving CREB1- miR-20a and target genes RRAGD, PIP4K2A, RHOC and CCND2 is altered between HMM and NHMM and is common between the 3 datasets. We have now begun to statistically ranks computationally predicted FFLs and develop a motif score to develop an integrated risk stratification model. In conclusion, FFLs form critical regulatory loops driving the functional behavior of MM cells. Analyzing the molecular impact of FFLs as a unit combining the aggregate impact of TF-miR and the target gene/s in MM will be instrumental in understanding the biology of the disease, developing clinically relevant integrated risk models, and translating basic research into targeted therapy. The ultimate goal is to develop strategies to regulate homeostatic control of these loops and overcome their oncogenic effects that drive the malignant phenotype.
Munshi:Celgene: Consultancy; Millenium: Consultancy; Merck: Consultancy; Onyx: Consultancy.
Asterisk with author names denotes non-ASH members.