Crumble: reference free lossy compression of sequence quality values

Abstract

Motivation: The bulk of space taken up by NGS sequencing CRAM files consists of per-base quality values. Most of these are unnecessary for variant calling, offering an opportunity for space saving. Results: On the CHM1+CHM13 test set, a 17 fold reduction in quality storage can be achieved while maintaining variant calling accuracy. Availability: Crumble is OpenSource and can be obtained from https://github.com/jkbonfield/crumble.