Accurate gene detection in noisy long reads

Reads produced by MinION technology^{25 (link)} are known to be noisy and contain frequent indel errors, a problem that also translates to assemblies derived from such long reads. In consequence, genes cannot be detected reliably on such DNA sequences. DIAMOND addresses this issue by providing frameshift alignments in translated search (blastx) mode. The protein sequences corresponding to all three reading frames of a strand are aligned simultaneously against the target sequence, allowing shifts in the reading frame at any position in the alignment, while incurring a user-defined score penalty (set using -F on the command line). The raw MinION reads and contigs up to the length of full bacterial chromosomes are supported as input in translated search mode, enabling gene discovery and annotation in the absence of known gene boundaries.