Inference-based accuracy of metagenome prediction tools varies across sample types and functional categories
Inference-based accuracy of metagenome prediction tools varies across sample types and functional categories
复制标题
DOI：
10.1186/s40168-020-00815-y
复制
发表时间：
2020-04-02
期刊：
影响因子：
15.5
通讯作者：
Fodor, Anthony A.
中科院分区：
文献类型：
--
作者：
Sun, Shan;Jones, Roshonda B.;Fodor, Anthony A.
背景：尽管测序成本最近有所下降，但与16S rRNA扩增子测序相比，鸟枪元基因组测序仍然更昂贵。已经开发出根据微生物群落的分类组成来预测其功能分布的方法。在这项研究中，我们评估了三种常用的元基因组预测工具(PICRUST、PICRUSt2和Tax4Fun)的性能，方法是比较不同环境下预测的功能基因谱与鸟枪式元基因组测序结果的差异丰度的重要性。结果我们选择了7个人类、非人类动物和环境(土壤)样本的数据集，这些数据集包含公开可用的16S rRNA和鸟枪元基因组序列。正如我们基于以前的文献所期望的那样，在预测的基因组成和用鸟枪式元基因组测序测量的基因相对丰度之间观察到了很强的Spearman相关性。然而，即使在不同样本之间的基因丰度发生变化时，这些强烈的相关性也会被保留下来。这表明，简单的相关系数是衡量超基因组预测工具表现的极不可靠的指标。作为另一种选择，我们将PICRUST、PICRUSt2和Tax4Fun预测的基因与每个数据集中与元数据相关的推理模型中测序的元基因组基因的性能进行了比较。通过这种方法，我们发现人类数据集的性能是合理的，元基因组预测工具在推断与“家务”功能相关的基因方面表现得更好。然而，当用于推理时，它们在人类数据集之外的性能急剧下降。结论PICRUST、PICRUSt2和Tax4Fun在人类样本以外的默认数据库中的推论作用可能是有限的，开发针对不同非人类和环境样本的基因预测工具是必要的。
Background Despite recent decreases in the cost of sequencing, shotgun metagenome sequencing remains more expensive compared with 16S rRNA amplicon sequencing. Methods have been developed to predict the functional profiles of microbial communities based on their taxonomic composition. In this study, we evaluated the performance of three commonly used metagenome prediction tools (PICRUSt, PICRUSt2, and Tax4Fun) by comparing the significance of the differential abundance of predicted functional gene profiles to those from shotgun metagenome sequencing across different environments. Results We selected 7 datasets of human, non-human animal, and environmental (soil) samples that have publicly available 16S rRNA and shotgun metagenome sequences. As we would expect based on previous literature, strong Spearman correlations were observed between predicted gene compositions and gene relative abundance measured with shotgun metagenome sequencing. However, these strong correlations were preserved even when the abundance of genes were permuted across samples. This suggests that simple correlation coefficient is a highly unreliable measure for the performance of metagenome prediction tools. As an alternative, we compared the performance of genes predicted with PICRUSt, PICRUSt2, and Tax4Fun to sequenced metagenome genes in inference models associated with metadata within each dataset. With this approach, we found reasonable performance for human datasets, with the metagenome prediction tools performing better for inference on genes related to "housekeeping" functions. However, their performance degraded sharply outside of human datasets when used for inference. Conclusion We conclude that the utility of PICRUSt, PICRUSt2, and Tax4Fun for inference with the default database is likely limited outside of human samples and that development of tools for gene prediction specific to different non-human and environmental samples is warranted.