How should I prepare and sequence samples for ChIP-seq?

04 Library Preparation and QC 05 Sequencing
If we prepare the sequencing libraries we require ChIP-seq DNA samples to be submitted after reversal of the cross-linking. Ideally, the fragment lengths should be between 100 and 300 bp, and preferably under 500 bp. The former will result in the tightest peaks.
For ChIP-seq it is common to start with DNA samples with concentrations too low to measure. Otherwise, the general DNA sample recommendations apply (buffer should be EB buffer or EBT: https://dnatech.genomecenter.ucdavis.edu/illumina-library-construction/) although more samples should be supplied if available. We highly recommend submitting the samples in low-retention tubes (e.g. Eppendorf LoBind).
General ChIP-seq recommendations:
  • The fragment lengths should be consistent and best be between 100 and 300 bp (up to 400 bp for the majority of molecules is acceptable).  Consistent fragment lengths can best be achieved on a Covaris style closed tube sonicator. We recommend avoiding probe sonicators.
  • Please make sure to run the input controls on a Bioanalyzer or agarose gel beforehand, and email us an image of these.
  • Sequence one "input control" per cell line/sample type.
  • Analyze at least two biological replicates.
  • We highly recommend verifying the enrichment of your regions of interest (e.g. promoter regions) vs. the control samples by qPCR, before submitting the samples for sequencing.
  • For highest accuracy data we can now generate sequencing libraries with UMI-bearing sequencing adapters. UMIs (Unique Molecular Identifiers) allow the accurate detection and removal of PCR duplicate reads. This approach is especially recommended for low-input samples. The first nine bases of the forward and reverse reads will contain UMI sequences.
The required read number per sample will vary from target to target. For the study of point source transcription factors the ENCODE project recommends analyzing at least 20 million (uniquely mapping) reads (https://genome.cshlp.org/content/22/9/1813.long#boxed-text-2). Depending on the quality of your preps, perhaps 75% of the reads can be expected to be uniquely mapping. ENCODE tends to err on the high side with their recommendations. Thus, about 20 million read pairs per sample should be acceptable, but this is likely the minimum number. Zhang et al. 2016 have studied the impact of the sequencing run types on ChIP-seq data analysis. Their data indicate that paired-end sequencing data provide significant advantages of single-end sequencing in ChIP-seq.

CUT&RUN sequencing might be a better alternative:

CUT&RUN sequencing (Skene & Henikoff 2017) is a faster protocol that for almost all applications is a more sensitive alternative requiring much lower cell numbers. CUT&RUN is suitable for studying histone modifications, transcription factors, and co-factors. In addition to lower input requirements CUT & RUN experiments also afford reduced read numbers (4 to 8 million read pairs per sample). References: Landt et al. 2012: ChIP-seq guidelines and practices of the ENCODE and modENCODE consortia. Genome Research 22: 1813-1831 Bailey et al. 2013: Practical Guidelines for the Comprehensive Analysis of ChIP-seq Data.  PLOS Computational Biology https://doi.org/10.1371/journal.pcbi.1003326 Skene & Henikoff 2017: An efficient targeted nuclease strategy for high-resolution mapping of DNA binding sites. https://doi.org/10.7554/eLife.21856 Zhang et al. 2016: Systematic evaluation of the impact of ChIP-seq read designs on genome coverage, peak identification, and allele-specific binding detection. BMC Bioinformatics volume 17, Article number: 96 MACS -- Model-based Analysis of ChIP-Seq https://taoliu.github.io/MACS/ https://hbctraining.github.io/Intro-to-ChIPseq/lessons/05_peak_calling_macs.html