The annelid Platynereis dumerilii is increasingly used as a model organism for developmental comparative studies due to its phylogenetic position and the accessibility of embryos that exhibit a stereotypic cleavage pattern and invariant cell lineages with predictable cell fates. To develop this unconventional model we established PdumBase, a comprehensive data base and intuitive online user interface based on stage specific transcriptomic data that allows genome wide identification of gene families contributing to particular biological processes during early developmental stages. One such important biological process is ciliogenesis, the formation of cilia, organelles associated with a variety of cellular roles such as motility, signaling, and sensory functions. However, knowledge of its multiple molecular components and regulatory mechanisms governing this dynamic process lags behind the functional understanding of cilia.

To close this gap and to highlight the versatility of the data encompassing PdumBase, we have developed an in silico guided identification pipeline for genes that contribute to the generation of a multiciliated cell type (MCC). Based on sequence similarity we identified orthologous P. dumerilii genes to the majority of the known ciliary genes described in other species. In addition, we validated their potential contribution to ciliogenesis through a differential expression analysis based on wild type vs. experimentally manipulated hyperciliated embryos. Our study revealed over 600 known ciliary genes to be significantly upregulated in treated embryos. These included genes encoding for well-known candidates in ciliogenesis such as dyneins, kinesis, microtubule organizers, basal body associated proteins, signaling proteins, and transcription factors, which were summarized into a set of high confidence core of ciliogenesis candidate genes.

To further associate genes that lack any annotation with ciliary activity, we developed DendroShiny, a computational approach to implicate potentially novel ciliary genes among poorly characterized transcripts based on rigorous statistical analysis and clustering of their co-expression patterns. DendroShiny achieves these goals by (1) clustering expression patterns of known genes and (2) using machine learning to determine expression features that allow for the classification of poorly characterized genes. Finally, our approach interactively displays the relationship of the identified clusters and their corresponding expression patterns, consequently facilitating the downstream analysis of transcriptomic data sets.

Taken together, our approach enables the identification of candidate and potentially novel ciliary genes despite the lack of an annotated genome and sets the ground for the elucidation of regulatory interactions between these candidate genes. This work hence represents a first step towards the generation of a comprehensive survey for ciliogenesis genes in Platynereis dumerilii.