ErrorX: automated error correction for immune repertoire sequencing datasets

MotivationRecent advances in DNA sequencing technology have allowed deep profiling of B- and T-cell receptor sequences on an unprecedented scale. However, sequencing errors pose a significant challenge in expanding the scope of these experiments. Errors can arise both by PCR during library preparation and by miscalled bases on the sequencing instrument itself. These errors compromise the validity of biological conclusions drawn from the data. ResultsTo address these concerns I have developed ErrorX, a software for automated error correction of B- and T-cell receptor NGS datasets. ErrorX uses deep learning to automatically identify bases that have a high probability of being erroneous. In benchmark studies, ErrorX reduced the overall error rate of public datasets by up to 36% with a false positive rate of 0.05% or less. Since ErrorX is a pure bioinformatics approach, it can be directly applied to any existing antibody or T-cell receptor sequencing datasets to infer sites of probable error without any changes in library preparation. AvailabilityErrorX is free for non-commercial use, with both a command-line interface and GUI available for Mac, Linux, and Windows operating systems, and full documentation available. Pre-compiled binaries are available at https://endeavorbio.com/downloads/.