The LINCS program offers a unique window on the mechanisms of response to small molecule and single gene perturbations by a diverse set of well-characterized cell lines. We propose to develop new algorithms for the analysis of the large-scale molecular profile data that will be made available by this program. These algorithms will help elucidate how response to small-molecule and biochemical perturbations is mediated by the genetic and molecular context of the cell and establish a predictive framework for the dissection of synergistic (i.e., non additive) perturbations. To achieve these goals, a new class of integrative methodologies is required, specifically tailored to the LINCS centers'data. Indeed, given the sparse, rather than genome-wide nature of these data, existing algorithms will not be directly applicable and will require either substantial customization or total rethinking. We have developed a repertoire of experimentally validated algorithms, such as ARACNe, for the dissection of transcriptional interactions (1, 2) and MINDy for post-translational interactions (3, 4) in mammalian cells. These are complements by methods for the integration of individual regulatory layers within multi-layer networks (5, 6), or interactomes for short. We have also pioneered algorithms (MARINa and IDEA) for interrogating interactomes to identify Master Regulator and Master Integrator genes, causally related to the presentation of pathologic (7-9) or physiologic phenotypes (5), and to elucidate compound Mechanism of Action (MoA) (6). Building on this extensive methodological base, we will introduce novel approaches, specifically tailored to the LINCS data, to (a) elucidate the MoA and activity of small molecule and RNAi/cDNA mediated perturbations (b) identify genes that mediate sensitivity or resistance to compound activity, and (c) identify gene-gene, gene-compound, and compound-compound interactions whose modulation can synergistically induce a desired endpoint phenotype (e.g., apoptosis, cell cycle arrest, loss of pluripotency, etc.). A specific emphasis of this proposal is on the use of LINCS in vitro signatures to predict such compound-related properties (MoA, activity, sensitivity, synergy, etc.) in vivo.
The premise ofthe LINCS program is that by understanding how and when a cell's phenotype is altered by specific stressors (as is the case in a diseased state), we can better understand the mechanisms underiying a given disease process. We are proposing an analytical framework that will allow matching compounds (including FDA approved drugs) to a disease tissue sample based on interactome analysis of disease- and comoound-based signatures, thus estahlishino an initial predictive framework for nersonalized medicine
| Mitrofanova, Antonina; Aytes, Alvaro; Zou, Min et al. (2015) Predicting Drug Response in Human Prostate Cancer from Preclinical Analysis of In Vivo Mouse Models. Cell Rep 12:2060-71 |
| Chiu, Hua-Sheng; Llobet-Navas, David; Yang, Xuerui et al. (2015) Cupid: simultaneous reconstruction of microRNA-target and ceRNA networks. Genome Res 25:257-67 |
| Woo, Jung Hoon; Shimoni, Yishai; Yang, Wan Seok et al. (2015) Elucidating Compound Mechanism of Action by Network Perturbation Analysis. Cell 162:441-451 |
| Repunte-Canonigo, Vez; Lefebvre, Celine; George, Olivier et al. (2014) Gene expression changes consistent with neuroAIDS and impaired working memory in HIV-1 transgenic rats. Mol Neurodegener 9:26 |
| Bansal, Mukesh; Yang, Jichen; Karan, Charles et al. (2014) A community computational challenge to predict the activity of pairs of compounds. Nat Biotechnol 32:1213-22 |
