Limits...
Identifying significant temporal variation in time course microarray data without replicates.

Billups SC, Neville MC, Rudolph M, Porter W, Schedin P - BMC Bioinformatics (2009)

Bottom Line: An important component of time course microarray studies is the identification of genes that demonstrate significant time-dependent variation in their expression levels.Until recently, available methods for performing such significance tests required replicates of individual time points.These results were confirmed in follow-up laboratory experiments.

View Article: PubMed Central - HTML - PubMed

Affiliation: Department of Mathematical and Statistical Sciences, University of Colorado, Denver, CO, USA. Stephen.Billups@ucdenver.edu

ABSTRACT

Background: An important component of time course microarray studies is the identification of genes that demonstrate significant time-dependent variation in their expression levels. Until recently, available methods for performing such significance tests required replicates of individual time points. This paper describes a replicate-free method that was developed as part of a study of the estrous cycle in the rat mammary gland in which no replicate data was collected.

Results: A temporal test statistic is proposed that is based on the degree to which data are smoothed when fit by a spline function. An algorithm is presented that uses this test statistic together with a false discovery rate method to identify genes whose expression profiles exhibit significant temporal variation. The algorithm is tested on simulated data, and is compared with another recently published replicate-free method. The simulated data consists both of genes with known temporal dependencies, and genes from a distribution. The proposed algorithm identifies a larger percentage of the time-dependent genes for a given false discovery rate. Use of the algorithm in a study of the estrous cycle in the rat mammary gland resulted in the identification of genes exhibiting distinct circadian variation. These results were confirmed in follow-up laboratory experiments.

Conclusion: The proposed algorithm provides a new approach for identifying expression profiles with significant temporal variation without relying on replicates. When compared with a recently published algorithm on simulated data, the proposed algorithm appears to identify a larger percentage of time-dependent genes for a given false discovery rate. The development of the algorithm was instrumental in revealing the presence of circadian variation in the virgin rat mammary gland during the estrous cycle.

Show MeSH
Estimated p-values (low freq. signals).
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC2682797&req=5

Figure 2: Estimated p-values (low freq. signals).

Mentions: In the first set of tests, we compared the p-values calculated by the two methods on 9 simulated data sets. Each data set corresponds to an underlying temporal variation with different frequency and amplitude, specified by parameters ω and α. (See (5) in Section 5.1 for details). The results are shown in Figures 1, 2, and 3. Figure 1 corresponds to the case, where there are no temporal dependencies in the data. The graphs in Figure 2 correspond to cases where there is a temporal dependency whose frequency is relatively low, and the graphs in Figure 3 correspond to higher frequency temporal dependencies. In each graph, the simulated genes are sorted in order of increasing p-values for Method 1.


Identifying significant temporal variation in time course microarray data without replicates.

Billups SC, Neville MC, Rudolph M, Porter W, Schedin P - BMC Bioinformatics (2009)

Estimated p-values (low freq. signals).
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC2682797&req=5

Figure 2: Estimated p-values (low freq. signals).
Mentions: In the first set of tests, we compared the p-values calculated by the two methods on 9 simulated data sets. Each data set corresponds to an underlying temporal variation with different frequency and amplitude, specified by parameters ω and α. (See (5) in Section 5.1 for details). The results are shown in Figures 1, 2, and 3. Figure 1 corresponds to the case, where there are no temporal dependencies in the data. The graphs in Figure 2 correspond to cases where there is a temporal dependency whose frequency is relatively low, and the graphs in Figure 3 correspond to higher frequency temporal dependencies. In each graph, the simulated genes are sorted in order of increasing p-values for Method 1.

Bottom Line: An important component of time course microarray studies is the identification of genes that demonstrate significant time-dependent variation in their expression levels.Until recently, available methods for performing such significance tests required replicates of individual time points.These results were confirmed in follow-up laboratory experiments.

View Article: PubMed Central - HTML - PubMed

Affiliation: Department of Mathematical and Statistical Sciences, University of Colorado, Denver, CO, USA. Stephen.Billups@ucdenver.edu

ABSTRACT

Background: An important component of time course microarray studies is the identification of genes that demonstrate significant time-dependent variation in their expression levels. Until recently, available methods for performing such significance tests required replicates of individual time points. This paper describes a replicate-free method that was developed as part of a study of the estrous cycle in the rat mammary gland in which no replicate data was collected.

Results: A temporal test statistic is proposed that is based on the degree to which data are smoothed when fit by a spline function. An algorithm is presented that uses this test statistic together with a false discovery rate method to identify genes whose expression profiles exhibit significant temporal variation. The algorithm is tested on simulated data, and is compared with another recently published replicate-free method. The simulated data consists both of genes with known temporal dependencies, and genes from a distribution. The proposed algorithm identifies a larger percentage of the time-dependent genes for a given false discovery rate. Use of the algorithm in a study of the estrous cycle in the rat mammary gland resulted in the identification of genes exhibiting distinct circadian variation. These results were confirmed in follow-up laboratory experiments.

Conclusion: The proposed algorithm provides a new approach for identifying expression profiles with significant temporal variation without relying on replicates. When compared with a recently published algorithm on simulated data, the proposed algorithm appears to identify a larger percentage of time-dependent genes for a given false discovery rate. The development of the algorithm was instrumental in revealing the presence of circadian variation in the virgin rat mammary gland during the estrous cycle.

Show MeSH