DNA Data Storage Research @ Imperial College

DNA data storage starts with translating any type of binary files (documents, movies, music, databases) to a binary representation, encoding the binary representation to sequences while incorporating error correction, and then synthesising the sequence representation to physical DNA molecules. The resulting molecules may be stored for years if not decades after which they are sequenced to recover the data. The resulting sequencing data is aligned, decoded and potential errors are fixed based on incorporated error correction codes to finally obtain the digital files encoded.

Our research focuses on all stages of the DNA data storage cycle:
      – Encoding, decoding and error correction for DNA data storage
      – Optimising synthesis processes for DNA data storage
      – Physical storage of DNA molecules for data storage
      – Sequencing for DNA storage based on Machine Learning/AI methods

DNA Data Storage

DNA Data Storage, the process of storing arbitrary binary data in DNA sequences is actively researched by us and the community. Its properties, primarily durability and compact form factor, make it an ideal storage media for archiving data for decades. Watch the video here to see what makes DNA an ideal storage media and how it all works.

Modelling DNA Storage as a Constrained Channel

Key to storing binary data in synthetic in DNA is the translation between the binary representation of digital data to the quaternary domain of DNA. This translation must adhere to constraints imposed by the synthesis and sequencing processes used to write and read respectively. A technological advance in either process changes the constraints and renders current encoding schemes obsolete. In this line of work we present a recipe for taking constraints and producing an appropriate encoding scheme. Such a mechanism allows moving the encoding in lockstep with the technological advances in the underlying processes. We further show a method to understand trade-offs in constraints for a given overhead of bits needed to meet such constraints