PUMA: PANDA Using MicroRNA Associations

Abstract

Conventional methods to analyze genomic data do not make use of the interplay between multiple 19 factors, such as between microRNAs (miRNAs) and the mRNA transcripts they regulate, and 20 thereby often fail to identify the cellular processes that are unique to specific tissues. We developed 21 PUMA (PANDA Using MicroRNA Associations), a computational tool that uses message passing to 22 integrate a prior network of miRNA target predictions with protein-protein interaction and target 23 gene co-expression information to model genome-wide gene regulation by miRNAs. We applied 24

Looking back at the first twenty years of genomics

“…The outputs from PANDA are condition‐specific regulatory networks that can be compared to understand condition‐specific differences in regulatory processes. PANDA has been extended to include additional regulatory factors such as miRNAs [21] and epigenetic factors [22], and a linear interpolation method, LIONESS, allows gene regulatory network models to be inferred for each individual sample analyzed in a study. Gene regulatory network modeling has been used to identify potential drug targets in ovarian cancer subtypes [23], identify regulatory changes as tissues are converted to cell lines [24], find network structures defining tissue‐specific functions in thirty‐eight different tissues [25], and comparing regulatory network drivers of sexual differences in twenty‐nine tissues [26] and in colon cancer [27].…”

“…Central to these models is the idea that gene expression (and downstream protein expression) is essential for determining phenotype. Indeed, methods that use gene-gene correlation measures to construct expression networks have proven their value in identifying co-expressed modules of genes representing functional groups that differ between phenotypes [20]. However, correlation in gene expression, while consistent with co-regulation, is not necessarily indicative of it.…”