Enrichment Constrained Time-Dependent Clustering Analysis for Finding Meaningful Temporal Transcription Modules

Jia Meng¹, Shou-Jiang Gao^2,3 and Yufei Huang^1,3*

¹Department of ECE, University of Texas at San Antonio, ²Department of Pediatrics, ³Greehey Children’s Cancer Research Institute, University of Texas Health Science Center at San Antonio

---

ABSTRACT

Motivation: Clustering is a popular data exploration technique widely used in microarray data analysis. When dealing with time series data, most conventional clustering algorithms, however, either use one-way clustering methods, which fail to consider the heterogeneity of temporary domain, or use two-way clustering methods that do not take into account the time dependency between samples, thus producing less informative results. Furthermore, enrichment analysis is often performed independent of and after clustering and such practice, though capable of revealing biological significant clusters, cannot guide the clustering to produce biologically significant result.

Result: We present a new enrichment constrained framework (ECF) coupled with a time-dependent iterative signature algorithm (TDISA), which, by applying a sliding time window to incorporate the time dependency of samples and imposing an enrichment constraint to parameters of clustering, allows supervised identification of temporal transcription modules (TTMs) that are biologically meaningful. Rigorous mathematical definitions of TTM as well as the enrichment constraint framework are also provided that serve as objective functions for retrieving biologically significant modules. We applied the Enrichment Constrained Time-Dependent Iterative Signature Algorithm (ECTDISA) to human gene expression time series data of Kaposi's sarcoma-associated herpesvirus (KSHV) infection of human primary endothelial cells; the result not only confirms known biological facts, but also reveals new insight into the molecular mechanism of KSHV infection.


Full text of the paper is available from the Journal's website here .

---

DATA AND CODE

• Download code and data
  You can get the data and Matlab code of the algorithm by clicking at the link.

---