• Explore more than 100 years of PNAS content!
  • Sign-up for PNAS eTOC Alerts

Reverse engineering dynamic temporal models of biological processes and their relationships

  1. Bud Mishrad,e,1
  1. aDepartment of Computer Science, Virginia Tech, Blacksburg, VA 24061;
  2. bDepartment of Mathematics, Virginia Tech, Blacksburg, VA 24061;
  3. cDepartment of Biochemistry, Virginia Tech, Blacksburg, VA 24061;
  4. dCourant Institute of Mathematical Sciences, New York University, New York, NY 10003;
  5. eNew York University (NYU) School of Medicine, New York, NY 10003; and
  6. fDipartimento di Informatica, Sistemistica e Comunicazione, Università degli Studi di Milano Bicocca, U14 Viale Sarca 336, I-20126 Milan, Italy

  1. Communicated by Michael H. Wigler, Cold Spring Harbor Laboratory, Cold Spring Harbor, NY, May 10, 2010 (received for review June 3, 2009)

  1. Fig. 1.
    View larger version:
    Fig. 1.

    Temporal process models reconstructed from segmentation algorithm. States are identified through the segmentation algorithm and edges are labeled by the experimental conditions under which the transitions are observed. (left) YCC. (right) YMC.

  2. Fig. 2.
    View larger version:
    Fig. 2.

    Combined temporal process model of the YCC, YMC, and exposure of yeast cells to HP and MD treatment.

  3. Fig. 3.
    View larger version:
    Fig. 3.

    Preview of results from segmenting the YCC dataset. Only one cycle is shown here. The YCC involves the staged coordination of several phases (M/G1, time points [1–3]; G1,S, time points [4–6]; and G2,M, time points [7–9]). (A) Mean expression profiles for each group of genes depict the changing emphasis across the three phases. Contingency tables capture the concerted grouping of genes within segments (B, first row) as well as the regroupings between segments (B, second row). Observe that the contingency tables in the first row involve significant enrichments whereas the tables in the second row approximate a uniform distribution. Gantt chart views (C) depict the temporal coordination of biological processes underlying the dataset. Only some of the enriched functions are displayed, for lack of space.

  4. Fig. 4.
    View larger version:
    Fig. 4.

    Segmentation resulting from the GOALIE analysis of transcriptional profiling datasets evaluating the rhythmical growth of S. cerevisiae (YMC1: diploid CEN.PK122, nutrient-limited conditions; YMC2: diploid IFO0233, not nutrient limited). The time line of each experiment is shown with each hash mark indicating a sampling point. GOALIE accurately determined the G1, S, and G2/M phases of the cell cycle, respectively. Note that the genes associated with each segment were culture and strain-dependent.

  5. Fig. 5.
    View larger version:
    Fig. 5.

    Segmentation resulting from the GOALIE analysis of a transcriptional profiling dataset evaluating the exposure of S. cerevisiae (BY8743) to HP (0.2 mM) and MD (2 mM). The time line of the experiment is shown; each hash mark indicates a sampling point, and the duration of the treatment is above the time line. GOALIE accurately assigned Segments I, II, and IV of the peroxide dataset to the times when the cells are predominantly in G1, S, and G2/M phases of the cell cycle, respectively. Segment III putatively represents the combined transition between phases of the cell cycle as well as the release from oxidative stress. Note the prevalence of genes associated with core metabolic processes including sulfur metabolism. GOALIE analysis of the MD treatment again resulted in the assignment of the cell cycle stages (I–III) as well as the G1 arrest.

Online Impact