AI 中文总结
本文提出IndelFreeAligner流式无间隙比对工具，采用索引与暴力双模式，通过蒙特卡洛模拟优化阈值，在基因组序列比对中速度与内存效率远超Bowtie1、BLAST+，为兆碱基级参考的比对任务提供高效方案。
AI 中文摘要
短序列与海量参考数据库的比对是现代基因组学的核心环节，但面临着显著的可扩展性挑战。传统比对工具依赖对参考序列进行耗时且高内存的预索引，这为小型查询集与兆碱基级数据的比对应用（如CRISPR间隔序列分析）造成了严重瓶颈。本文提出IndelFreeAligner，这是一款消除了预处理瓶颈的流式无间隙比对工具，它支持两种模式：适用于大型查询集的索引模式，以及针对小型查询集追求极致速度的暴力模式。通过实时处理参考序列，IndelFreeAligner支持用户指定的错配阈值，最高可达完整查询序列长度，且内存使用量与参考序列总大小无关。其新颖的MinHitsCalculator组件采用蒙特卡洛模拟来确定索引模式下的自适应种子命中阈值。与Bowtie1和BLAST+的基准测试显示：IndelFreeAligner将单个查询与4 Gbp参考序列比对耗时1.7秒，而Bowtie1（含最优线程数下的索引构建）耗时17分钟，速度提升607倍；针对RefSeq细菌数据库（560 GB压缩），IndelFreeAligner用8 GB内存完成10个查询的搜索耗时12分钟，而BLAST+仅构建数据库就需3小时17分钟，查询时则需506 GB内存。索引模式在4个错配下实现0假阴性（8-16个错配时的映射率为99.84%-99.90%），暴力模式设计上为穷尽式比对。IndelFreeAligner为此前计算上难以实现的比对任务提供了可扩展且高效的解决方案，它作为BBTools（Bushnell, 2014）的一部分以开源形式发布。
英文摘要
The comparison of short sequences to massive reference databases is a cornerstone of modern genomics, but it presents a significant scalability challenge. Traditional alignment tools rely on time- and memory-intensive pre-indexing of the reference, creating a substantial bottleneck for applications involving small query sets against terabase-scale data, such as CRISPR spacer analysis. Here we present IndelFreeAligner, a streaming, indel-free alignment tool that eliminates the preprocessing bottleneck. It operates in two modes: an indexed mode for larger query sets and a brute-force mode for maximum speed on small query sets. By processing reference sequences on-the-fly, IndelFreeAligner supports user-specified mismatch thresholds up to the full query length and maintains memory usage independent of total reference size. A novel MinHitsCalculator component uses Monte Carlo simulation to determine adaptive seed-hit thresholds for indexed mode. Benchmarks against Bowtie1 and BLAST+ demonstrate that IndelFreeAligner aligns a single query against a 4 Gbp reference in 1.7 seconds versus 17 minutes for Bowtie1 (including index construction at optimal thread count), a 607-fold speedup. Against RefSeq Bacteria (560 GB compressed), IndelFreeAligner completes a 10-query search in 12 minutes using 8 GB of RAM, while BLAST+ requires 3 hours and 17 minutes to build its database alone and 506 GB of RAM to query it. Indexed mode achieves 0% false negatives through 4 substitutions (99.84-99.90% mapped at 8-16); brute-force mode is exhaustive by design. IndelFreeAligner provides a scalable and efficient solution for alignment tasks that were previously computationally prohibitive. It is distributed open-source as part of BBTools (Bushnell, 2014).
Comments9 pages, 2 figures, 7 tables
