DUDE-Seq: Fast Universal Denoising of Nucleotide Sequences

We consider the correction of errors from nucleotide sequences produced by next-generation sequencing. The error rate in reads has been increasing with the shift of focus of mainstream sequencers from accuracy to throughput. Denoising in high-throughput sequencing is thus becoming a crucial component for boosting the reliability of downstream analyses. Our methodology, named DUDE-Seq, is derived from a general setting of reconstructing finite-valued source data corrupted by a discrete memoryless channel and provides an effective means for correcting substitution and homopolymer indel errors, the two major types of sequencing errors in most high-throughput sequencing platforms. Our experimental studies with real and simulated data sets suggest that the proposed DUDE-Seq outperforms existing alternatives in terms of error-correction capabilities, time efficiency, as boosting the reliability of downstream analyses. Further, DUDE-Seq is universally applicable across different sequencing platforms and analysis pipelines by a simple update of the noise model. [availability: http://data.snu.ac.kr/pub/dude-seq]