Euler circuits and DNA sequencing by hybridization

Sequencing by hybridization is a method of reconstructing a long DNA string - that is, figuring out its nucleotide sequence - from knowledge of its short substrings. Unique reconstruction is not always possible, and the goal of this paper is to study the number of reconstructions of a random string. For a given string, the number of reconstructions is determined by the pattern of repeated substrings; in an appropriate limit substrings will occur at most twice, so the pattern of repeats is given by a pairing: a string of length 2n in which each symbol occurs twice. A pairing induces a 2-in, 2-out graph, whose directed edges are defined by successive symbols of the pairing - for example the pairing ABBCAC induces the graph with edges AB, BB, BC, and so forth - and the number of reconstructions is simply the number of Euler circuits in this 2-in, 2-out graph. The original problem is thus transformed into one about pairings: to find the number fk(n) of n-symbol pairings having k Euler circuits. We show how to compute this function, in closed form, for any fixed k, and we present the functions explicitly for k=1,...,9. The key is a decomposition theorem: the Euler circuit number of a pairing is the product of the circuit numbers of component sub-pairings. These components come from connected components of the interlace graph , which has the pairing\u27s symbols as vertices, and edges when symbols are interlaced . (A and B are interlaced if the pairing has the form ABAB or BABA.) We carry these results back to the original question about DNA strings, and provide a total variation distance upper bound for the approximation error. We perform an asymptotic enumeration of 2-in, 2-out digraphs to show that, for a typical random n-pairing, the number of Euler circuits is of order no smaller than 2n/n, and the expected number is asymptotically at least e-1/22n-1/n. Since any n-pairing has at most 2n-1 Euler circuits, this pinpoints the exponential growth rate. © 2000 Elsevier Science B.V