MOTIVATION: While biological systems operated from a common genome can be conserved in various ways, they can also manifest highly diverse dynamics and functions. This is because the same set of genes can interact differentially across specific molecular contexts. For example, differential gene interactions give rise to various stages of morphogenesis during cerebellar development. However, after over a decade of efforts toward reverse engineering biological networks from high-throughput omic data, gene networks of most organisms remain sketchy. This hindrance has motivated us to develop comparative modeling to highlight conserved and differential gene interactions across experimental conditions, without reconstructing complete gene networks first. RESULTS: We established a comparative dynamical system modeling (CDSM) approach to identify conserved and differential interactions across molecular contexts. In CDSM, interactions are represented by ordinary differential equations and compared across conditions through statistical heterogeneity and homogeneity tests. CDSM demonstrated a consistent superiority over differential correlation and reconstruct-then-compare in simulation studies. We exploited CDSM to elucidate gene interactions important for cellular processes poorly understood during mouse cerebellar development. We generated hypotheses on 66 differential genetic interactions involved in expansion of the external granule layer. These interactions are implicated in cell cycle, differentiation, apoptosis and morphogenesis. Additional 1639 differential interactions among gene clusters were also identified when we compared gene interactions during the presence of Rhombic lip versus the presence of distinct internal granule layer. Moreover, compared with differential correlation and reconstruct-then-compare, CDSM makes fewer assumptions on data and thus is applicable to a wider range of biological assays. AVAILABILITY: Source code in C++ and R is available for non-commercial organizations upon request from the corresponding author. The cerebellum gene expression dataset used in this article is available upon request from the Goldowitz lab (dang@cmmt.ubc.ca, http://grits.dglab.org/). CONTACT: joemsong@cs.nmsu.edu SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.
MOTIVATION: While biological systems operated from a common genome can be conserved in various ways, they can also manifest highly diverse dynamics and functions. This is because the same set of genes can interact differentially across specific molecular contexts. For example, differential gene interactions give rise to various stages of morphogenesis during cerebellar development. However, after over a decade of efforts toward reverse engineering biological networks from high-throughput omic data, gene networks of most organisms remain sketchy. This hindrance has motivated us to develop comparative modeling to highlight conserved and differential gene interactions across experimental conditions, without reconstructing complete gene networks first. RESULTS: We established a comparative dynamical system modeling (CDSM) approach to identify conserved and differential interactions across molecular contexts. In CDSM, interactions are represented by ordinary differential equations and compared across conditions through statistical heterogeneity and homogeneity tests. CDSM demonstrated a consistent superiority over differential correlation and reconstruct-then-compare in simulation studies. We exploited CDSM to elucidate gene interactions important for cellular processes poorly understood during mouse cerebellar development. We generated hypotheses on 66 differential genetic interactions involved in expansion of the external granule layer. These interactions are implicated in cell cycle, differentiation, apoptosis and morphogenesis. Additional 1639 differential interactions among gene clusters were also identified when we compared gene interactions during the presence of Rhombic lip versus the presence of distinct internal granule layer. Moreover, compared with differential correlation and reconstruct-then-compare, CDSM makes fewer assumptions on data and thus is applicable to a wider range of biological assays. AVAILABILITY: Source code in C++ and R is available for non-commercial organizations upon request from the corresponding author. The cerebellum gene expression dataset used in this article is available upon request from the Goldowitz lab (dang@cmmt.ubc.ca, http://grits.dglab.org/). CONTACT: joemsong@cs.nmsu.edu SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.
Authors: Nicolas Le Novère; Benjamin Bornstein; Alexander Broicher; Mélanie Courtot; Marco Donizelli; Harish Dharuri; Lu Li; Herbert Sauro; Maria Schilstra; Bruce Shapiro; Jacky L Snoep; Michael Hucka Journal: Nucleic Acids Res Date: 2006-01-01 Impact factor: 16.971
Authors: Debra L Fulton; Saravanan Sundararajan; Gwenael Badis; Timothy R Hughes; Wyeth W Wasserman; Jared C Roach; Rob Sladek Journal: Genome Biol Date: 2009-03-12 Impact factor: 13.583
Authors: Mingzhou Song; Yang Zhang; Alexia J Katzaroff; Bruce A Edgar; Laura Buttitta Journal: Nucleic Acids Res Date: 2014-01-29 Impact factor: 16.971
Authors: Travis B Cotton; Hien H Nguyen; Joseph I Said; Zhengyu Ouyang; Jinfa Zhang; Mingzhou Song Journal: Sci Rep Date: 2015-04-28 Impact factor: 4.379