A linear-time algorithm for finding the longest segment which scores above a given threshold

This paper describes a linear-time algorithm that finds the longest stretch in a sequence of real numbers (“scores”) in which the sum exceeds an input parameter. The algorithm also solves the problem of finding the longest interval in which the average of the scores is above a fixed threshold. The problem originates from molecular sequence analysis: for instance, the algorithm can be employed to identify long GC-rich regions in DNA sequences. The algorithm can also be used to trim low-quality ends of shotgun sequences in a preprocessing step of whole-genome assembly.