Mantis: A Fast, Small, and Exact Large-Scale Sequence-Search Index

a Motivation. Sequence-level searches on large collections of RNA-seq experiments, such as the NIH Sequence Read Archive (SRA), would enable one to ask many questions about the expression or variation of a given transcript in a population. Bloom filter-based indexes and variants, such as the Sequence Bloom Tree, have been proposed in the past to solve this problem. However, these approaches suffer from fundamental limitations of the Bloom filter, resulting in slow build and query times, less-than-optimal space usage, and large numbers of false positives. Results. This paper introduces Mantis, a space-efficient data structure that can be used to index thousands of rawread experiments and facilitate large-scale sequence searches on those experiments. Mantis uses counting quotient filters instead of Bloom filters, enabling rapid index builds and queries, small indexes, and exact results, i.e., no false positives or negatives. Furthermore, Mantis is also a colored de Bruijn graph representation, so it supports fast graph traversal and other topological analyses in addition to large-scale sequence-level searches. In our performance evaluation, index construction with Mantis is 4.4× faster and yields a 20% smaller index than the state-of-the-art split sequence Bloom tree (SSBT). For queries, Mantis is 6×–108× faster than SSBT and has no false positives or false negatives. For example, Mantis was able to search for all 200,400 known human transcripts in an index of 2652 human blood, breast, and brain RNA-seq experiments in one hour and 22 minutes; SBT took close to 4 days and AllSomeSBT took about eight hours. Mantis is written in C++11 and is available at https://github.com/splatlab/mantis.