Molecular-level similarity search brings computing to DNA data storage.

Area of Science:

• Biotechnology
• Bioinformatics
• Data Storage

Background:

• Digital storage demand is increasing, driving the need for dense and durable storage solutions.
• Synthetic DNA offers a promising alternative to traditional storage media due to its density and longevity.
• Current DNA data retrieval methods primarily support exact file access via keys, limiting complex data operations.

Purpose of the Study:

• To develop and demonstrate a method for performing similarity search on data stored in synthetic DNA.
• To enable content-based retrieval of images from a large DNA database without prior knowledge of file names.

Main Methods:

• Encoding images into DNA sequences using a learned image-to-sequence mapping.
• Designing hybridization probes as queries for similarity search.
• Executing similarity search on a DNA database containing 1.6 million images.

Main Results:

• Successfully demonstrated similarity search over a DNA-based database of 1.6 million images.
• The developed molecular approach achieved performance comparable to state-of-the-art in silico similarity search algorithms.
• The image-to-sequence encoding was crucial for preferential binding of queries to visually similar target sequences.

Conclusions:

• This work presents a novel capability for complex data computations, specifically similarity search, within DNA data storage systems.
• The findings pave the way for more sophisticated data analysis and retrieval directly from DNA archives.
• Synthetic DNA storage can support advanced functionalities beyond simple file storage and retrieval.