Recommended coverage depth for ChIP Seq (H3K9me3 )

Question

According to ENCODE recommendations

For broad-peak histone experiments, each replicate should have 45 million usable fragments

According to my understanding, a usable fragment means a uniquely mapped read / pair of reads.

However

H3K9me3 is an exception as it is enriched in repetitive regions of the genome. ... Tissues and primary cells should have 45 million total mapped reads per replicate.

I fail to grasp Encode's language.

• Does "total mapped reads" mean uniquely mapped reads?
• If so, why is 45 million "an exception" ? It is the same as the other broad peaks.

Any ideas where else can I look for an explanation?

Answer 1

The ENCODE Data Coordination Center, that developed and maintains the ENCODE portal, have implemented scripts validating adherence to the standards the ENCODE Consortium agreed upon when evaluating ChIP-seq analysis results. For any ChIP-seq experiment, except for H3K9me3, the adherence to the read depth standard is assessed looking on the number of "usable fragemnts" (https://www.encodeproject.org/data-standards/terms/#read-depth) (see answer by James Hawley). For H3K9me3 the read depth is assessed not by looking on usable fragments, but by looking on the total number of mapped reads (if it is a PE sequencing run, the number of mapped read pairs). It is an exception, because the number of mapped reads, especially for H3K9me3 is substantially higher than the number of usable fragments (where the duplicated and low mapping quality reads are filtered out). The decision to apply different read depth standard in case of H3K9me3 came up due to the logic explained here (https://www.encodeproject.org/chip-seq/histone/), briefly: because the H3K9me3 mark is enriched in repetitive regions of the genome, the number of usable fragments (post-filtering) is not expected to be as high as other marks (45M), despite high-quality libraries and sufficient sequencing depth. Instead of marking these experiments as failing read depth standard, the script checks the total number of mapped reads (or read pairs for PE sequencing runs) and makes sure it is above the read depth threshold (45M).

Answer 2

In the initial description for "usable fragments", ENCODE has the following (emphasis mine):

Usable fragment – A fragment is defined as the sequencing output corresponding to one location in the genome. If single-ended sequencing is performed, one read is considered a fragment. If paired-ended sequencing is performed, one pair of reads is considered a fragment. Fragments are considered usable if they pass the various filters in the ChIP-seq uniform processing pipelines. Used to evaluate ChIP-seq data.

Aligning reads to repetitive regions is difficult because they don't give unique alignments. Reads from these regions will often have poor alignment scores and may not pass alignment QC checks. Reads may also be counted as multi-mapped, and often methods call for uniquely mapped reads, which again, may not be true for a large number of reads from H3K9me3 ChIP-seq.

So for this particular histone mark, ENCODE loosens the definition for what you should count towards a minimum sequencing depth threshold.