Abstract
Motivation Advances in high-throughput sequencing technologies has resulted in the generation of high coverage sequencing datasets. Assembling such datasets is challenging for current assemblers in terms of computational resources. It has been shown previously that a large part of these datasets is redundant and can be removed, without affecting assembly quality. But determining which reads to eliminate is a challenge as it might have an impact on the quality of the assembly produced. A de Bruijn graph (DBG), which is the base for many de novo assemblers, uses k-mers from reads as nodes and generate assemblies by traversing the nodes. Hence, removal of reads might result in loss of k-mers forming connections between genomic regions.
Results This work presents ORNA, a set multicover based read normalization algorithm. It focuses on the fact that each connection in the DBG is k + 1-mer connecting two k-mers. It selects the minimum number of reads which is required to retain all the k + 1-mers from the original dataset and hence retains all the connections from the original graph. Different datasets were normalized using ORNA and assembled using different DBG based assemblers. ORNA was also compared against well established read normalization algorithms was mostly found to be better.
Conclusion ORNA allows read normalization without losing connectivity information of the original DBG. It is fast and memory efficient. Together with available error correction approaches, OR able to provide high quality assembly. This is a step forward in making the assembly of high coverage dataset easier and computationally inexpensive.
ORNA is available at https://github.com/SchulzLab/ORNA
