A Fast and Symmetric DUST Implementation to Mask Low-Complexity DNA Sequences

The DUST module has been used within BLAST for many years to mask low-complexity sequences. In this paper, we present a new implementation of the DUST module that uses the same function to assign a complexity score to a sequence, but uses a different rule by which high-scoring sequences are masked. The new rule masks every nucleotide masked by the old rule and occasionally masks more. The new masking rule corrects two related deficiencies with the old rule. First, the new rule is symmetric with respect to reversing the sequence. Second, the new rule is not context sensitive; the decision to mask a subsequence does not depend on what sequences flank it. The new implementation is at least four times faster than the old on the human genome. We show that both the percentage of additional bases masked and the effect on MegaBLAST outputs are very small.