Several years ago, we presented the BLIND method which is able to determine the temporal order of transcriptomes in large datasets, thus allowing for the massive collection of samples, to be ordered at the analysis stage:

Anavy L, Levin M, Khair S, Nakanishi N, Valverde SF, Degnan BM, and Yanai I. BLIND ordering of large-scale transcriptomic developmental time-courses. Development 2014 141:1161-6.

We demonstrated that this method can correctly recover the annotated order of embryonic transcriptome time-courses that have been previously published. Furthermore, to test BLIND on a large-scale dataset we created a high-resolution embryonic time-course for the sponge Amphimedon queenslandica. We collected 59 embryos and examined each both morphologically and using RNA-Seq, resulting in a developmental time-course more detailed than any published one. We found that BLIND was able to correctly identify both the order of the samples – as well as the direction of time – without any annotation whatsoever beyond the gene expression data. BLIND is thus particularly useful to developmental biologists as they examine organisms with asynchronous stages or those with stages that are difficult to identify morphologically.

Here is a link to the BLIND Matlab script.

papers