proTRAC - a software for probabilistic piRNA cluster detection, visualization and analysis
Institute of Anthropology, Johannes Gutenberg-University Mainz, Colonel-Kleinmann-Weg 2, 55099 Mainz, Germany
BMC Bioinformatics 2012, 13:5 doi:10.1186/1471-2105-13-5Published: 10 January 2012
Throughout the metazoan lineage, typically gonadal expressed Piwi proteins and their guiding piRNAs (~26-32nt in length) form a protective mechanism of RNA interference directed against the propagation of transposable elements (TEs). Most piRNAs are generated from genomic piRNA clusters. Annotation of experimentally obtained piRNAs from small RNA/cDNA-libraries and detection of genomic piRNA clusters are crucial for a thorough understanding of the still enigmatic piRNA pathway, especially in an evolutionary context. Currently, detection of piRNA clusters relies on bioinformatics rather than detection and sequencing of primary piRNA cluster transcripts and the stringency of the methods applied in different studies differs considerably. Additionally, not all important piRNA cluster characteristics were taken into account during bioinformatic processing. Depending on the applied method this can lead to: i) an accidentally underrepresentation of TE related piRNAs, ii) overlook duplicated clusters harboring few or no single-copy loci and iii) false positive annotation of clusters that are in fact just accumulations of multi-copy loci corresponding to frequently mapped reads, but are not transcribed to piRNA precursors.
We developed a software which detects and analyses piRNA clusters (proTRAC, probabilistic TRacking and Analysis of Clusters) based on quantifiable deviations from a hypothetical uniform distribution regarding the decisive piRNA cluster characteristics. We used piRNA sequences from human, macaque, mouse and rat to identify piRNA clusters in the respective species with proTRAC and compared the obtained results with piRNA cluster annotation from piRNABank and the results generated by different hitherto applied methods.
proTRAC identified clusters not annotated at piRNABank and rejected annotated clusters based on the absence of important features like strand asymmetry. We further show, that proTRAC detects clusters that are passed over if a minimum number of single-copy piRNA loci are required and that proTRAC assigns more sequence reads per cluster since it does not preclude frequently mapped reads from the analysis.
With proTRAC we provide a reliable tool for detection, visualization and analysis of piRNA clusters. Detected clusters are well supported by comprehensible probabilistic parameters and retain a maximum amount of information, thus overcoming the present conflict of sensitivity and specificity in piRNA cluster detection.