The option of omic data created from international consortia, aswell as from world-wide laboratories, offers the chance both to answer long-standing questions in biomedicine/molecular biology also to formulate novel hypotheses to check. McLeay et al., 2012). Analogously, gene manifestation changes have already been correlated to changes of TF bindings and chromatin marks (Althammer et al., 2012; Klein et al., 2014). Generally, gene manifestation can be expected utilizing a limited amount of examples (in specific circumstances). On the contrary, inferring huge GRNs could be reached just MK-2206 2HCl distributor using many high-throughput datasets, as with Gerstein et al. (2012). Nevertheless, some networks could be simpler than expected and may rely on a minimal number of factors and interactions. Dunn et al. (2014) recently identified a minimal set of components (12 TFs and 16 interactions) sufficient to explain the self-renewal of ES cells. In terms of potential impact on human genetics, we highlight the following considerations. Cell differentiation is accompanied by globaland localchromatin changes, leading to the silencing of pluripotency genes and lineage-specific gene activation (Chen and Dent, 2014). In this regard, multi-omic integration and single-cell omics can be used to explain and to potentially control differentiation and to explore heterogeneity of cells in development and disease (Comes et al., 2013; Macaulay and Voet, 2014). Understanding such mechanisms will significantly improve MK-2206 2HCl distributor the treatment of human genetic diseases, particularly of cancer. Indeed, epigeneticunlike geneticmodifications are reversible, and modulating epi-marks through up/down-regulation of histone methyltransferases can affect gene expression and tissue-specific alternative splicing (Luco et al., 2010, 2011). By correcting the aberrant distribution of epi-marks, we may in turn control pathologic changes in gene expression (Schenk et al., 2012). In this regard, the proper identification of aberrant epigenetic regulators in tumors is of major interest. The final objective is to identify new therapeutic targets and to develop novel molecules (TF binding prediction could be used as surrogate information, in absence of binding data. Open in a separate window Figure 2 One of the key-points in the integration process is the way in which the epigenetic and transcriptional signals are transformed into a statistical model that relates a response vector Y (i.e., gene expression) with a set of predictors, represented by a matrix X (i.e., epigenetic signatures). (A) A scheme showing gene transcription, and the molecular factors involved (TFs and HMs), is illustrated in the upper part. (B) Different models have been proposed to build the so-called gene to epigenetic signature matrix X. Naive models proposed to use a binary matrix to integrate epigenetic signatures with gene expression. Therefore, 0/1 values were used to annotate and associate a given TF or HM to a specific gene according to a proximity measure MK-2206 2HCl distributor between the peak and/or the enriched region and TSS of the corresponding gene. More advanced models, such as the one from Ouyang et al. (2009), proposed to use a weighed sum of peaks around the TSS. In this way it is possible to tune the strength of the binding and the distance from the TSS in a continuous way. Along the same direction, Sikora-Wohlfeld MK-2206 2HCl distributor et al. (2013) compared several other measures to build X. All such approaches share the idea that TSC2 matrix X is built with respect to the position of the TSSs (or using reads in a window around the TSSs) by collapsing each epigenetic MK-2206 2HCl distributor feature into.