适用于QIIME2的UNITE 9.0分类器及训练方法

在使用 QIIME2 分析 ITS 数据时，需要注释降噪得到的代表序列，而注释需要输入所参考的数据库。

QIIME2官方给的说法是In our experience, fungal ITS classifiers trained on the UNITE reference database do NOT benefit from extracting/trimming reads to primer sites. We recommend training UNITE classifiers on the full reference sequences. Furthermore, we recommend the “developer” sequences (located within the QIIME-compatible release download) because the standard versions of the sequences have already been trimmed to the ITS region (excluding portions of flanking rRNA genes that may be present in amplicons generated with standard ITS primers).

那就自己训练一个好咯。

1下载并导入序列到qiime

在UNITE下载链接[1]选择合适的版本下载数据库，下载解压后一共有这6个文件。

# 将序列导入qiime
qiime tools import \
--type FeatureData[Sequence] \
--input-path Database/UNITE_qiime/sh_refs_qiime_ver9_99_25.07.2023.fasta \
--output-path unite-ver9-seqs_99_25.07.2023.qza

2导入分类学文件

qiime tools import \
--type FeatureData[Taxonomy] \
--input-path Database/UNITE_qiime/sh_taxonomy_qiime_ver9_99_25.07.2023.txt \
--output-path unite-ver9-taxonomy_99_25.07.2023.qza \
--input-format HeaderlessTSVTaxonomyFormat

3训练分类器

这个步骤比较耗时（5h+），也比较占用内存（最多70G）。

nohup qiime feature-classifier fit-classifier-naive-bayes \
--i-reference-reads unite-ver9-seqs_99_25.07.2023.qza \
--i-reference-taxonomy unite-ver9-taxonomy_99_25.07.2023.qza \
--o-classifier unite-ver9-99-classifier-25.07.2023.qza &

网上也有其他人发布自己训练的分类器如pre-trained UNITE 9.0 classifiers for QIIME 2023.9 (and older!)[2]

PS：本想设置免费分享unite-ver9-99-classifier-25.07.2023.qza的链接，无奈阿里不让分享qza文件，若有需要可打赏后在后台私信。

参考资料

[1]

UNITE下载链接: https://unite.ut.ee/repository.php

[2]

pre-trained UNITE 9.0 classifiers for QIIME 2023.9 (and older!): https://forum.qiime2.org/t/pre-trained-unite-9-0-classifiers-for-qiime-2023-9-and-older/24140/1