Naren Ramakrishnan and Richard F. HelmDepartments of Computer Science and BiochemistryVirginia Tech, Blacksburg, VA 24061

What we do How we do it Some discussion of biological results

Outline

What we doFrom multivariate time course data, reconstruct (i) stages, (ii) transitions between stages, and (iii) properties that hold in each stage

What are temporal models?

Kripke structures

Temporal compartmentalization Yeast store carbohydrate and burn it in late G1

phase of the cell cycle: this superimposes a metabolic cycle over the cell cycle

Sequence of oxidative and reductive phases◦ O2 is used up to burn carbohydrates during oxidative phase◦ O2 levels increase and carbohydrate is stored in the

reductive phase Interplay between the YCC and YMC lead to

compartmentalization of biological processes in time◦ Cell cycle : reductive phase◦ Glycolysis and respiration occur at different times

Combined Kripke model

Segment the time course Identify biological processes enriched in

each segment Convert segmentation to a Kripke structure Merge Kripke structures into a dynamic

temporal model

How do we construct temporal models?

Time series segmentation How do we identify “breakpoints” in the

time course? Intuition: A breakpoint is where there is

significant reorganization of clusters around segment boundaries.

Modeling Cluster DynamicsA Cluster Dynamics Approach3602 genes, 3 clusters

Modeling Cluster DynamicsA Cluster Dynamics Approach3602 genes, 3 clusters

Algorithm optimizes for this!

A second example (YCC)

Gantt charts of YMC

Gantt chart of HP treatment

Gantt chart of MD treatmentGantt chart of MD treatment

Combined Kripke model again

Thank you