A comparison of seed-and-extend techniques in modern DNA read alignment algorithms

Abstract

Abstract-DNA read alignment is a major step in genome analysis. However, as DNA reads continue to become longer, new approaches need to be developed to effectively use these longer reads in the alignment process. Modern aligners commonly use a two-step approach for read alignment: 1. seeding, 2. extension. In this paper, we have investigated various seeding and extension techniques used in modern DNA read alignment algorithms to find the best seeding and extension combinations. We developed an open source generic DNA read aligner that can be used to compare the alignment accuracy and total execution time of different combinations of seeding and extension algorithms. For extension, our results show that local alignment is the best extension approach, achieving up to 3.6x more accuracy than other extension techniques, for longer reads. For seeding, if BLAST-like seed extension is used, the best seeding approach is identifying all SMEMs in the DNA read (e.g., approach used by BWA-MEM). This combination is up to 6x more accurate than other seeding techniques, for longer reads. With local alignment, we observed that the seeding technique does not impact the alignment accuracy. Furthermore, we showed that an optimized implementation of local alignment using vector instructions, enabling 4.5x speedup, makes it the fastest of all extension techniques. Overall, we show that using local alignment with non-overlapping maximal exact matching seeds is the best seeding-extension combination due to its high accuracy and higher potential for optimization/acceleration for future DNA reads.

A performant bridge between fixed-size and variable-size seeding

“…Subfigure ii) shows the situation of i) for 2-step 3-mers, a subset of 3-mers that contains every second 3-mer merely. The first and last nucleotides of s are not covered by 2-step 3-mers and must be discovered by an additional extension For a reference R and query Q, the set of all MEMs is defined, in accordance with [6], as follows: For all pairs of reference-query positions (x, y), with R[x] = Q[y], 0 ≤ x < |R|, 0 ≤ y <|Q|, we get a MEM (q, r, l) as follows: Let i be the minimal value such that R…”

“…Most high-throughput read aligners [1][2][3][4][5] perform the following three steps: seeding [6,7], seed processing (e.g. chaining) [8,9] and dynamic programming [10,11].…”

“…Variable-size seeding relies on maximal exact matches (MEMs) or subsets of them. A MEM [6] is a perfect match between query and reference that cannot be extended further in either direction. MEMs can be computed directly via some form of full-text search index as e.g.…”

A performant bridge between fixed-size and variable-size seeding

“…Via the algorithmic bridge proposed here (Alg. 1), it is possible to define a mapping between -mers and MEMs (maximal exact matches, see [6]). The mapping is surjective, because every MEM has at least one corresponding -mer.…”

“…Computation of SMEMs out of MEMs.Algorithm 2a shows the Pseudocode for extraction of SMEMs of MEMs using a sorting followed by a single sweep over all MEMs. In order to prove the correctness of the algorithm, we characterize SMEMs (in accordance to[6,15]) as follows: A SMEM is a MEM that is not enclosed by another MEM on the query. The above algorithm identifies all enclosed MEMs by iterating over them ordered by their start positions on the query.…”