Alignment and mapping methodology influence transcript abundance estimation
DOI(url): https://doi.org/10.1101/657874
发表日期：June 03, 2019
关键点
不对比对方法对转录本定量的影响有哪些（读完感觉是给 Salmon 最近一次升级写的软文）
参考意义
使用 RNA-seq 数据进行转录本定量的准确性取决于许多因素，比如比对的方法和所采用的定量模型。虽然有不少文章已经讲过定量模型的重要性，但比较各种比对方法对定量准确度的影响并没有那么受关注。作者在这篇文章中研究了比对方法对定量准确性以及对差异基因表达分析的影响。
即使定量模型本身不变，选择不同的比对方法，或使用不同的参数对定量的影响有时可能很大并影响下游分析。作者也强调当评估过于注重模拟数据时，这些影响可能会被我们忽视，因为在模拟数据中，比对这一步往往比实验获得的样本更简单。文章讨论了用于定量目的的最佳比对方法，同时也引入了一种新的混合比对方法，称为 selective alignment(SA)。
文章中，作者选择了三种比对策略：
- unspliced alignment of RNA-seq reads directly to the transcriptome
- spliced alignment of RNA-seq reads to the annotated genome (with subsequent projection to the transcriptome)
- (unspliced) lightweight mapping (quasi-mapping) of the RNA-seq reads directly to the transcriptome
具体的比对方法：
- Bowtie2 – Alignment with Bowtie2 to the target transcriptome and allowing alignments with indels, followed by quantification using Salmon in alignment mode.
- Bowtie2 strict – Alignment with Bowtie2 to the target transcriptome and disallowing alignments with indels (i.e. using the same parameters as those used by RSEM), followed by quantification using Salmon in alignment mode.
- Bowtie2 RSEM – Alignment with Bowtie2 to the target transcriptome and disallowing alignments with indels, followed by quantification using RSEM.
- STAR – Alignment with STAR to the target genome (aided with the GTF annotation of the transcriptome) and projected to the transcriptome allowing alignments with indels and soft clipping, followed by quantification using Salmon in alignment mode.
- STAR strict – Alignment with STAR to the target genome (aided with the GTF annotation of the transcriptome) and projected to the transcriptome and disallowing alignments with indels or soft clipping, followed by quantification using Salmon in alignment mode.
- STAR RSEM – Alignment with STAR to the target genome (aided with the GTF annotation of the transcriptome) and projected to the transcriptome and disallowing alignments with indels or soft clipping, followed by quantification using RSEM.
- quasi – Quasi-mapping directly to the target transcriptome, coupled with quantification using Salmon in non-alignment mode.
- SA– Selective alignment directly to the target transcriptome and a set of decoy sequences, coupled with quantification using Salmon in non-alignment mode.
相关内容
# For indexing, we use the following extra command line arguments, along with the regular indexing and
threads parameters:
STAR --genomeFastaFiles <fasta file> --sjdbGTFfile <gtf file> --sjdbOverhang 100
Bowtie2 default
salmon -k 23 --keepDuplicates
kallisto -k 23
# For quantification, we use the following extra command line, \
# along with regular index and threads, with each tools we compare against:
SA --mimicBT2 --useEM
quasi --rangeFactorization 4 --discardOrphansQuasi --useEM
Bowtie2 --sensitive -k 200 -X 1000 --no-discordant --no-mixed
Bowtie2 strict --sensitive --dpad 0 --gbar 99999999 --mp 1,1 \
--np 1 --score-min L,0,-0.1 --no-mixed --no-discordant -k 200 -I 1 -X 1000
Bowtie2 RSEM --sensitive --dpad 0 --gbar 99999999 --mp 1,1 \
--np 1 --score-min L,0,-0.1 --no-mixed --no-discordant -k 200 -I 1 -X 1000
STAR --outFilterType BySJout --alignSJoverhangMin 8 --outFilterMultimapNmax 20 \
--alignSJDBoverhangMin 1 --outFilterMismatchNmax 999 \
--outFilterMismatchNoverReadLmax 0.04 --alignIntronMin 20 --alignIntronMax 1000000 \
--alignMatesGapMax 1000000 --readFilesCommand zcat --outSAMtype BAM Unsorted \
--quantMode TranscriptomeSAM --outSAMattributes NH HI AS NM MD \
--quantTranscriptomeBan Singleend
STAR strict --outFilterType BySJout --alignSJoverhangMin 8 --outFilterMultimapNmax \
20 --alignSJDBoverhangMin 1 --outFilterMismatchNmax 999 \
--outFilterMismatchNoverReadLmax 0.04 --alignIntronMin 20 --alignIntronMax 1000000 \
--alignMatesGapMax 1000000 --readFilesCommand zcat --outSAMtype BAM Unsorted \
--quantMode TranscriptomeSAM --outSAMattributes NH HI AS NM MD \
--quantTranscriptomeBan IndelSoftclipSingleend
STAR RSEM --outFilterType BySJout --alignSJoverhangMin 8 --outFilterMultimapNmax \
20 --alignSJDBoverhangMin 1 --outFilterMismatchNmax 999 \
--outFilterMismatchNoverReadLmax 0.04 --alignIntronMin 20 --alignIntronMax 1000000 \
--alignMatesGapMax 1000000 --readFilesCommand zcat --outSAMtype BAM Unsorted \
--quantMode TranscriptomeSAM --outSAMattributes NH HI AS NM MD \
--quantTranscriptomeBan IndelSoftclipSingleend \
RSEM default
kallisto default or --rf-stranded as appropriate
这里所谓的 SA 模式其实就是在 salmon 最近升级中加入的一个参数，具体可以了解 官方说明
A practical guide to methods controlling false discoveries in computational biology
DOI(url): https://doi.org/10.1186/s13059-019-1716-1
发表日期：4 June 2019
关键点
在数据分析的过程中如何更好的控制 false discoveries
参考意义
以下是 8 中可用的 FDR-controlling methods ，其中 IHW 和 BL 是考虑了协变量的现代方法。
