Quantifying molecular bias in DNA data storage

DNA has recently emerged as an attractive medium for archival data storage. Recent work has demonstrated proof-of-principle prototype systems; however, very uneven (biased) sequencing coverage has been reported, which indicates inefficiencies in the storage process. Deviations from the average coverage in the sequence copy distribution can either cause wasteful provisioning in sequencing or excessive number of missing sequences. Here, we use millions of unique sequences from a DNA-based digital data archival system to study the oligonucleotide copy unevenness problem and show that the two paramount sources of bias are the synthesis and amplification (PCR) processes. Based on these findings, we develop a statistical model for each molecular process as well as the overall process. We further use our model to explore the trade-offs between synthesis bias, storage physical density, logical redundancy, and sequencing redundancy, providing insights for engineering efficient, robust DNA data storage systems.