技术服务描述

QC Report

Report QC介绍基本信息，说明：

• 报告发布时间

• 报告类型atac

• 比对基因组使用的是什么，如mm10,hg38,rn6

• 比对软件bowtie2

• 测序类型为双端测序（paired end）

• peak caller 软件为 macs2

Alignment quality metrics (比对质量检测)

SAMstat (raw unfiltered BAM) (原始比对质量检测)

原始数据比对基因组过程的质控：

• Total Reads，为全部reads数。

• duplicate Reads，为重复reads数，需要去除。

• Mapped Reads，为mapping到基因组上的reads数。

• % Mapped Reads，为reads的mapping率。
  （大于90%非常好，大于80%较好，较低时可考虑是否是实验的特殊处理或其他污染导致。）

• Properly Paired Reads，正确匹配的双端reads。

• % Properly Paired Reads，reads 双端匹配率。

• With itself，双端匹配到基因组上的reads。

• singleton，只有单端匹配到基因组上的reads。

• % singleton，只有单端的匹配率。

• Diff.Chroms，没有匹配到基因组上的数据。

Marking duplicates (filtered BAM) (重复数据标记检测)

Filtered out (samtools view -F 1804):

• read unmapped (0x4)

• mate unmapped (0x8, for paired-end)

• not primary alignment (0x100)

• read fails platform/vendor quality checks (0x200)

• read is PCR or optical duplicate (0x400)

说明：
Duplicate Reads是由于PCR过度扩增，导致的完全一致的reads，这些reads在分析时需要去除。利用samtools过滤掉的数据包括：

• unmapped的reads（单端/双端）

• 没有唯一匹配的reads

• 本身质量差的reads

• PCR重复产生的duplicate reads

SAMstat (filtered/deduped BAM) (过滤掉线粒体reads及重复reads后的质量检测)

Filtered and duplicates removed

说明：

• duplicate reads，对过滤后的reads进行质控。表格说明同上。

Sequence quality metrics (filtered/deduped BAM) (测序质量检测)

Open chromatin assays are known to have significant GC bias. Please take this            into consideration as necessary.

说明：

• 红色线条为每条序列在基因组特定位置的均一化覆盖度，

• 蓝色线条为每条序列的平均碱基的质量

• 绿色线条为每条碱基的GC含量，正常一般分布在50%左右，

每条序列的平均碱基的质量参考说明：

Phred分值	不正确的碱基识别	碱基正确识别率	Q.sorce
10	1/10	90%	Q10
20	1/100	99%	Q20
30	1/1000	99.9%	Q30
40	1/10000	99.99%	Q40

Library complexity quality metrics (文库复杂度质量检测)

Library complexity (filtered non-mito BAM) (文库复杂度)

Mitochondrial reads are filtered out by default.        The non-redundant fraction (NRF) is the fraction of non-redundant mapped reads        in a dataset; it is the ratio between the number of positions in the genome        that uniquely mapped reads map to and the total number of uniquely mappable        reads. The NRF should be > 0.8. The PBC1 is the ratio of genomic locations        with EXACTLY one read pair over the genomic locations with AT LEAST one read        pair. PBC1 is the primary measure, and the PBC1 should be close to 1.        Provisionally 0-0.5 is severe bottlenecking, 0.5-0.8 is moderate bottlenecking,        0.8-0.9 is mild bottlenecking, and 0.9-1.0 is no bottlenecking. The PBC2 is        the ratio of genomic locations with EXACTLY one read pair over the genomic        locations with EXACTLY two read pairs. The PBC2 should be significantly        greater than 1. See more details at                the ENCODE portal standard for ChIP-Seq pipeline

NRF (non redundant fraction)
       PBC1 (PCR Bottleneck coefficient 1)
       PBC2 (PCR Bottleneck coefficient 2)
       PBC1 is the primary measure. Provisionally

• 0-0.5 is severe bottlenecking

• 0.5-0.8 is moderate bottlenecking

• 0.8-0.9 is mild bottlenecking

• 0.9-1.0 is no bottlenecking

说明：

• Total Fragments:总片段数

• Distinct Fragments:可以唯一匹配到基因组上的片段数

• One Reads：只有一个reads匹配到特定位置的片段数

• Two Read：有两个reads匹配的同一位置的片段数

• NRF = Distinct/Total

• PBC1 = OneRead/Distinct

• PBC2 = OneRead/TwoRead

文库复杂度的指标由三个值决定，NRF、PBC1、PBC2，其中最重要的为PBC1的数值

ENCODE对文库复杂度的要求如下，

PBC1	PBC2	Bottlenecking level	NRF	Complexity	Flag colors
< 0.7	<1	Severe	< 0.7	Concerning	Orange
0.7 <= PBC1 <= 0.9	1 <= PBC2 <= 3	Moderate	0.7 <= NRF <= 0.9	Acceptable	Yellow
> 0.9	>3	None	> 0.9	ldeal	None

Replication quality metrics (重复性质量检测)

IDR (Irreproducible Discovery Rate) plots (不可重复发现率)
（可选，仅在有至少2组重复(rep1,rep2,...)时计算）

IDR主要比较两个样本之间的重复性，先对peaks进行排序，然后进行重复性的分析，得到IDR得分。IDR<0.05，认为有重复性，黑色表示，IDR>=0.05认为没有重复性，红色表示。

• 一般对两重复样本进行IDR检测（rep1 vs rep2），检测两样本之间的重复性。

• 单样本的虚拟样本间也可进行IDR检测（rep1-pr1 vs rep1-pr2），验证IDR检测的准确性。

• 样本混样成pool进行IDR检测（pool-pr1 VS pool-pr2）

Reproducibility QC and peak detection statistics（重复性QC及peak统计分析）

Reproducibility QC

• N1: Replicate 1 self-consistent peaks (comparing two pseudoreplicates generated by subsampling Rep1 reads)

• N2: Replicate 2 self-consistent peaks (comparing two pseudoreplicates generated by subsampling Rep2 reads)

• Ni: Replicate i self-consistent peaks (comparing two pseudoreplicates generated by subsampling RepX reads)

• Nt: True Replicate consistent peaks (comparing true replicates Rep1 vs Rep2)

• Np: Pooled-pseudoreplicate consistent peaks (comparing two pseudoreplicates generated by subsampling pooled reads from Rep1 and Rep2)

• Self-consistency Ratio: max(N1,N2) / min (N1,N2)

• Rescue Ratio: max(Np,Nt) / min (Np,Nt)

• Reproducibility Test: If Self-consistency Ratio >2 AND Rescue Ratio > 2, then 'Fail' else 'Pass'

说明：

• N1: replicate1自我虚拟peak的一致性peak

• N2: replicate2自我虚拟peak的一致性peak

• Ni: replicatei自我虚拟peak的一致性peak

• Nt: 真实一致性的peak

• Np: 整体的数据（pool）自我虚拟peak的一致性peak

• N optimal: 虚拟peak的一致性peak

• N conservation: 真实peak的一致性peak

• Self-consistency Ratio:   max(N1,N2) / min (N1,N2)

• Rescue Ratio:    max(Np,Nt) / min (Np,Nt)

• Reproducibility Test: 如果Self-consistency Ratio >2 AND Rescue Ratio > 2, 则失败（Failed）,否则通过（pass）

ENCODE对以上指标的检测标准如下：

Self-consistency Ratio	Rescue Ratio	Resulting Data Status	Flag colors
Less than 2	Less than 2	Ideal	None
Less than 2	Greater than 2	Acceptable	Yellow
Greater than 2	Less than 2	Acceptable	Yellow
Greater than 2	Greater than 2	Concerning	Orange

• IDR 一致的peak 最好大于70,000 ，不少于50,000 为可接受

Number of raw peaks （原始peaks数）

           Top 500000 raw peaks from macs2 with p-val threshold 0.01

说明：

• 原始peak数最好不少于150,000，不少于100,000为可接受

Peak calling statistics （peak calling统计）

Peak region size （peak长度范围）

说明：

• 图形横坐标为region size，纵坐标为peak数。

• Rep1与rep2显示的是原始peaks（去除duplicate及线粒体基因组）

• Idr_opt显示的是经过IDR优化后的peaks

• Overlap_opt显示经过overlap优化后的peaks

Enrichment / Signal-to-noise ratio

Strand cross-correlation measures (trimmed/filtered SE BAM) (链互相关检测)

Performed on subsampled (15000000) reads mapped from FASTQs that are trimmed to 50.            Such FASTQ trimming and subsampling reads are for cross-corrleation analysis only.            Untrimmed FASTQs are used for all the other analyses.

NOTE1: For SE datasets, reads from replicates are randomly subsampled to 15000000.
           NOTE2: For PE datasets, the first end (R1) of each read-pair is selected and trimmed to 50 the reads are then randomly subsampled to 15000000.

• Normalized strand cross-correlation coefficient (NSC) = col9 in outFile

• Relative strand cross-correlation coefficient (RSC) = col10 in outFile

• Estimated fragment length = col3 in outFile, take the top value

NSC与RSC的算法如下：

从flitered BAM文件中虚拟一个含有25,000,000个reads的副本，并对该副本进行链互相关分析。链互相关性分析主要显示+，-链reads的相似性。由于正负reads的peak中心有偏移，因此，结果中一般会出现两个peak，一个是标准peak，一个是虚拟peak（ Phantom Peak）。如下图所示

Jensen-Shannon distance (filtered/deduped BAM)(JS散度检测)

说明：

• JS散度是检测两样本之间的离散程度，也就是检测两样本之间的相似性。

Peak enrichment

Fraction of reads in peaks (FRiP) (peaks占总reads数的比值)

FRiP for macs2 raw peaks

FRiP for overlap peaks

FRiP for IDR peaks

For macs2 raw peaks:

• repX: Peak from true replicate X

• repX-prY: Peak from Yth pseudoreplicates from replicate X

• pooled: Peak from pooled true replicates (pool of rep1, rep2, ...)

• pooled-pr1: Peak from 1st pooled pseudo replicate (pool of rep1-pr1, rep2-pr1, ...)

• pooled-pr2: Peak from 2nd pooled pseudo replicate (pool of rep1-pr2, rep2-pr2, ...)

           For overlap/IDR peaks:

• repX_vs_repY: Comparing two peaks from true replicates X and Y

• repX-pr1_vs_repX-pr2: Comparing two peaks from both pseudoreplicates from replicate X

• pooled-pr1_vs_pooled-pr2: Comparing two peaks from 1st and 2nd pooled pseudo replicates

说明：

• FRiP指的是fraction of all mapped reads that fall into peak regions. 代表peak区域内reads的比例，peak区域内的reads是抗体富集的序列，其他区域的序列是背景噪声。 只有mapping到基因组上的reads才会进行后续分析，所以直接用peak区域reads数目除以所有mapping基因组的数目，就得到了FRiP Score。

• Encode团队发现FRip Score和peak的个数存在正相关关系。 对于不同的转录因子chip数据，peak的个数不同，总体的趋势来看，peak的个数越多，FRiP Score的值越大。在Encode的chip数据集中，有80%左右的FRiP Score值都超过了1%，所以将1%看做是一个衡量chip实验的软标准。 之所以称之为软标准，是因为FRiP score低于1%并不能直接说明chip实验是失败的，高于1%也不能直接说明chip实验就是成功的。在ZNF274和RNA III型聚合酶的chip实验中，peak的个数很少，FRiP score的值也小于1%；在CTCF等转录因子实验中，FRiP score的值又远大于1%。

• 所以说FRiP score值是和特定的转录因子或者组蛋白修饰相关，在研究特定转录因子时，参考别人已经发表的数据中的FRiP Score作为衡量标准是更好的，在没有参考的情况下，可以用1%的阈值来作为一个软标准，为chip实验的质量提供一个参考。当peak个数的大于1000时，1%这个标准的效果还是比较好的，有相当大的参考意义。