Solving the coverage depth problem in DNA data storage using LT codes

The amount of data that needs to be stored in the digital world is increasing exponentially on a global scale. However, the storage media currently in use struggle to keep up with this rapid growth. This situation has led to the search for new storage technologies. Recent studies have shown that DNA can be used as a data storage medium. Nevertheless, its high cost remains a major disadvantage. In this thesis, we address the coverage depth problem in DNA data storage. DNA storage systems are subject not only to insertion, deletion, and substitution errors, but also to constraints such as GC content and strand length. Therefore, reading a single DNA strand is not sufficient to recover the original data. The coverage depth problem refers to the number of DNA strands from which at least ttt copies must be obtained in order to reconstruct the original information. Two studies are conducted in this thesis to address the coverage depth problem. The first study formulates an optimization problem to achieve the best decoding performance of LT codes under three cost factors: synthesis, sequencing, and error probability. The results of this study are presented in comparison with the coverage depth of MDS codes, and it is shown that MDS codes incur lower cost. The second study proposes a new algorithm, named DupLeTh, which is based on a two-phase LT decoding approach and aims to reduce the required coverage depth. The performance of the proposed DupLeTh algorithm is compared with that of MDS and systematic MDS codes. For data types such as video, audio, and images, lossy reconstruction can be sufficient to obtain meaningful and functional information. When losses are taken into account, the LT-code-based DupLeTh algorithm achieves better results than both MDS and systematic MDS codes.