Approximate all-pairs suffix/prefix overlaps

Finding approximate overlaps is the first phase of many sequence assembly methods. Given a set of strings of total length n and an error-rate epsilon, the goal is to find, for all-pairs of strings, their suffix/prefix matches (overlaps) that are within edit distance k = inverted right perpendicular epsilon linverted left perpendicular, where e is the length of the overlap. We propose a new solution for this problem based on backward backtracking (Lam, et al., 2008) and suffix filters (Karkkainen and Na, 2008). Our technique uses nH(k) + o(n log sigma) + r log r bits of space, where H-k is the k-th order entropy and sigma the alphabet size. In practice, it is more scalable in terms of space, and comparable in terms of time, than q-gram filters (Rasmussen, et al., 2006). Our method is also easy to parallelize and scales up to millions of DNA reads. (C) 2012 Elsevier Inc. All rights reserved.