用单分子测序（single-molecule sequencing）和局部敏感哈希（locality-sensitive hashing）来组装大型基因组

Assembling large genomes with single-molecule sequencing and locality-sensitive hashing

好好读读，算法系列的好文章！

Assembling large genomes with single-molecule sequencing and locality-sensitive hashing - NATURE BIOTECHNOLOGY

marbl/MHAP  - Github

MinHash Alignment Process (MHAP): a probabilistic sequence overlap algorithm.  - ReadTheDocs

PacificBiosciences/blasr – Github

Frequently Asked Questions: Data File Formats

BLASR M4 format - MHAP的输出格式

摘要

---

单分子实时测序技术（SMRT）常被用于完成微生物基因组，但是可用的组装方法还没有规模化应用到大型基因组上。

我们引入了MinHash Alignment Process (MHAP)来比对高噪音、长的reads，使用概率学和locality-sensitive hashing。

集成了MHAP的Celera Assembler使得 reference-grade的de novo组装变为可能（…）。

组装的结果高度的连续，包含了染色体臂、close persistent gaps的完整解决方案。

我们的D. melanogaster组装结果揭示了先前未知的异染色质和端粒序列，也组装了低复杂性的CHM1，从而填补了人类GRCh38的gap。

使用MHAP、CA和SMRT可以denovo出近乎完整的真核基因组，准确率达到99.99%。

前言

---

The primary bottleneck of long-read assembly has been the sensitive all-versus-all alignment required to determine overlapping read pairs.

长reads组装的主要瓶颈是两两比对的敏感性，用于决定reads对的overlap。

本文提供了一种概率算法，可以高效地检测出高错误长reads之间的overlap。

MHAP uses a dimensionality reduction technique named MinHash to create a more compact representation of sequencing reads.

MHAP使用了MinHash 的降维技术来创建了测序reads的更加紧凑的表示形式。

MinHash 最初是开发用来检测不同网页之间的相似度，它将文本或字符串减少到了一系列的fingerprints，称为sketch。

结果

---

MinHash alignment filtering

MHAP overlapping performance

SMRT sequencing and assembly

De novo human assembly using long reads

Assembly validation and repeat resolution

Improved telomere assemblies

讨论

---

待续~