ATD Collaborative Research: New theorems and algorithms for comprehensive analysis of metagenomic data via statistical phylogenetics

目前的全宏基因组分析工具主要基于序列相似性（组装，BLAST）和分类学（“分箱”方法）;虽然有用，但这些方法具有以下限制。组装方法需要读段重叠，因此仅重建混合样品中最丰富的生物体。序列相似性方法如BLAST不能将读段与祖先生物联系起来，也不能表明突变的进化意义。分类学方法过于粗糙，无法反映可能表征生物威胁的微妙DNA序列变化。研究人员建议通过发展方法的理论基础来克服这些局限性，以便：重建环境样品中DNA的细胞区室化，即使当读段计数很小时，通过与现存基因组进行系统发育比较来检测宏基因组样品内的合成基因组和定向进化的证据，检测给定其位置或观察时间的异常遗传物质的组合，从统计学上区分微生物群落组成中有意义的变化和噪音，即使这些变化发生在低于使用现有方法可检测到的水平上。基因工程的工具掌握在许多国家的科学家手中;这些工具可用于合成生物武器。防止这些武器造成的伤亡取决于能否迅速发现和识别这些武器。虽然高通量DNA测序可用于监测生物威胁，但目前用于分析其产生的大量信息的工具不足以对威胁风险进行统计分析。通过对进化信号的统计分析，生物防御监测方法可以直接从“宏基因组”数据中检测遗传异常和威胁：环境样本的高通量鸟枪测序数据。

Current whole-metagenome analysis tools are primarily based on sequence similarity (assembly, BLAST) and taxonomies ("binning" approaches); while useful, these approaches have the following limitations. Assembly methods require read overlap and thus only reconstruct the most abundant organisms in a mixed sample. Sequence similarity approaches such as BLAST cannot relate reads to ancestral organisms and do not indicate the evolutionary significance of mutations. Taxonomic methods are too coarse to reflect the subtle DNA sequence changes that may characterize a biological threat. The investigators propose to overcome these limitations by developing the theoretical underpinnings of methods to: reconstruct the cellular compartmentalization of DNA in environmental samples, even when read counts are small, detect synthetic genomes and evidence of directed evolution within a metagenomic sample by performing a phylogenetic comparison with extant genomes, detect combinations of genetic material that are anomalous given their location or time of observation, statistically distinguish meaningful shifts in microbial community composition from noise, even when those shifts happen at a level below that detectable using currently available methods.The tools of genetic engineering are in the hands of scientists of many countries; these tools can be used to synthesize biological weapons. Prevention of casualties from these weapons depends on their prompt detection and identification. Although high-throughput DNA sequencing could be used to monitor biological threats, the currently available tools for analyzing the wealth of information it generates are insufficient to statistically analyze threat risk. A biodefense monitoring approach informed by a statistical analysis of evolutionary signal could yield a means to detect genetic anomalies and threats directly from "metagenomic" data: high throughput shotgun sequencing data from environmental samples.