Saccharomyces cerevisiae Cell Cycle Example

Data set: cell-cycle - G1- S/G2.tab

The study by Spellman et al. (1998) identified 800 genes whose transcript levels vary periodically within the cell cycle. Genes were classified into several groups, based on their correlation to genes known to be expressed in corresponding phases of the cell cycle.

Genes from G1 and S/G2 phase groups were used as an example in this supplement. Key feature of G1 genes is peak expression in the mid-G1 phase (time point cdc15_150), approximately 10 minutes before the onset of the S phase. Similarly, the S/G2 group was determined by high expression in the time points corresponding to change form the S (at time point cdc15_80) to G2 phase (at time point cdc15_90) of the cell cycle.

Figure 1 shows gene expression profiles of the two phase groups: G1 and S/G2. Top row indicates cell cycle phase (Spellman et al., 1998). Here one can see that genes from phase G1 are highly expressed at time points cdc15_150 and cdc28_20, which is the G1 phase (colored green). Similarly, the S/G2 group is highly expressed at time point cdc_90, which is the time of transition from S (colored blue, time point cdc_80) to G2 (colored red).

Figure 1. Heat maps of the two phase groups of genes.

VizRank rediscovers the three most important attributes that determine the G1 phase (i.e. cdc15_140, cdc15_150 and cdc28_20), and the cdc15_90 attribute that is important for S/G2. These are exactly the attributes Spellman et al. (1998) used to classify genes into the two groups. In Figure 2.a one can see a clear distinction between the two groups.

Interestingly, VizRank finds that using an attribute (alpha49) originally assigned to determine a third group (i.e. phase M genes), we can get better discrimination between the two classes.

(a)		(b)
Figure 2. (a) Radviz projection using five attributes to best discriminate between two classes, with score 98.58. (b) Not including the alpha49 attribute results in a worse score of  95.50.

A top score scatterplot found by VizRank (Figure 3.a, with score 94.16) reconfirms that we can discriminate between groups G1 and S/G2 based on attributes cdc15_90 and cdc28_20. Those are same time points as in the radviz Figure 2.a, where three additional time points increase the discriminative quality between classes (radviz has score 98.58).

Figure 3.b is the result of a bad choice of attributes used to generate a scatterplot. There is no clear distinction between the two groups of genes.

(a)		(b)
Figure 3. (a) Best scatterplot projection found by VizRank, with score 94.16. (b) Example of a low-ranked scatterplot, with score 55.85.

References:

Spellman, P.T., et al. (1998) Comprehensive identification of cell cycle-regulated genes of the yeast Saccharomyces cerevisiae by microarray hybridization. Mol Biol Cell,  9(12), 3273-97.