Toward a dynamic threshold for quality-score distortion in reference-based alignment

The intrinsic high entropy sequence metadata, known as quality scores, are largely the cause of the substantial size of sequence data files. Yet, there is no consensus on a viable reduction of the resolution of the quality score scale, arguably because of collateral side effects. In this paper we leverage on the penalty functions of HISAT2 aligner to rebin the quality score scale in such a way as to avoid any impact on sequence alignment, identifying alongside a distortion threshold for \"safe\" quality score representation. We tested our findings on whole-genome and RNA-seq data, and contrasted the results with three methods for lossy compression of the quality scores.