In this folder hierarchy, we hub all QC Reports from HiChIP that we can get our hands on to facilitate comparisons. We’ve pre-categorized current reports into libraries that look good and bad. Additionally, we’ve done some analyses to determine some noteworthy features of the HiChIP experiment/sequencing run (see below). Have more reports to contribute? We’d greatly appreciate it if you submitted a pull request or emailed Caleb with your report!
When planning a HiChIP experiment and sequencing run, we consider three key factors: 1) cellular input, 2) read length, and 3) read count.
A paper published in late 2017 in Nature Genetics used
K27ac for HiChIP. See the loop call QC reports from hichipper here:
Compiled HTML for 1 million 5 million 10 million
While each sample looks mostly successful in terms of long range interactions, we observe significant heterogeneity in ChIP efficacies as the % of reads in anchors varies considerably within sample batches.
The original HiChIP paper only sequenced using 75 bp PE reads for all samples. To infer the impact of read length, we sequencing a high quality library at 100 bp PE reads and then trimmed these reads to 75 bp. From our analyses, this causes a 5-6% reduction in mappable interactions and ~12-13% reduction in reads in loops. Depending on the availability of sequencing technology, we suggest using 75 bp reads with more depth may be a more economical experimental strategy.
Three samples were created from random downsampling of the GM12878 HiChIP experiment with 1 million, 500,000 and 250,000 reads. As the proportion of long range interactions remains relatively constant, we suggest that a relatively small number of reads may be sufficient to determine library quality before performing a deeper sequencing run. This comparison is just a reorganization of data in the Good/Mumbach_etal/ folder. The original SRR files shown here include SRR3467183 (10 million), SRR3467185 (5 million), and SRR3467187 (1 million).