The Sierra Platinum Service for generating peak-calls for replicated ChIP-seq experiments

ChIP-seq has become an important high throughput technique for analyzing protein–DNA interaction. It is routinely employed for identifying transcription factor binding sites and for determining chromatin states by virtue of immunoprecipitation of nucleosomes that exhibit histones with specific chemical modifications. The basic principle of ChIP-seq is the specific enrichment of immunoprecipitated DNA-protein aggregates, from which—after sequencing the DNA component—the genomic location of the DNA–protein interactions of interest are inferred. The key step in the analysis of ChIP-seq data is peak-calling, that is, the determination of those genomic regions in which immunoprecipitated DNA is significantly enriched relative to “empty” control samples [1].

Since the introduction of ChIP-seq, several peak callers were published, e.g., MACS [2], PeakSeq [3], and csaw [4]. However, most of them handle only one pair of experiment and control or perform a differential peak-calling between two experiments. Their performances were reviewed by Wilbanks and Facciotti [5] and by Koohy et al. [6]. Tools for multi-replicate peak-calling were developed only recently. Among them, Sierra Platinum [7] combines multiple ChIP-seq experiments in a single peak calling process and thus makes full use of the information supplied by replicates. It provides extensive visualization options to guide the user through the evaluation of the experiments. This enables the user to inspect the replicate’s quality in various ways, and to enhance the peak-calling quality by weighting or excluding individual samples. The GUI of Sierra Platinum further facilitates immediately comparing the impact of different parameter settings. Compared to other currently available peak-calling tools, Sierra Platinum performs best with respect to recall and false discovery rate regardless of the data quality [7].

The computational efforts for ChIP-seq analysis require hardware that may not be available in labs without dedicated bioinformatics infrastructure due to the size of the input files and the complexity of the algorithms that combine multiple samples. To overcome this limitation, Sierra Platinum Service provides access to the full functionality of Sierra Platinum by providing a web-based service hosted at sierra.sca-ds.de. In addition, we provide a convenient docker image for easy deployment of private instances of the service, e.g., for institution-wide use.

The SPS architecture is currently designed as a split infrastructure. The registration and validation of a new user is handled by a web interface whereas the data upload, data processing, and data presentation is implemented within a Java GUI application. Therefore, the user needs an up-to-date Java installation on his client. Moreover, the data upload of the input files can take a long time depending on the user’s internet connection and the input size. If the user has a very slow upload rate, the maximal runtime of the service may exceed before finishing the upload. Further, the user needs a valid email address for the registration process.

Currently, the SPS instance at sierra.sca-ds.de is able to compute five jobs simultaneously. If necessary, the service can be extended easily with more slots since it is running on cluster system.