Towards Computation, Space, and Data Efficiency in de novo DNA Assembly: A Novel Algorithmic Framework

We consider the problem of de novo DNA sequencing from shot gun data, wherein an underlying (unknown) DNA sequence is to be reconstructed from several short substrings of the sequence. We propose a de novo assembly algorithm which requires only the minimum amount of data and is efficient with respect to space and computation. We design the algorithm from an information theoretic perspective of using minimum amount of data. The key idea to achieve space and computational efficiency is to break the procedure into two phases, an online and an offline phase. We remark that this can serve as an evidence of the feasibility of using an information-theoretic perspective to guide practical algorithmic design in DNA sequencing. Preliminary work on extending this algorithmic framework to more realistic settings is also reported.