Comprehensive, integrated, and phased whole-genome analysis of the primary ENCODE cell line K562

Abstract

K562 is one of the most widely used human cell lines in biomedical research. It is one of three tier-one cell lines of ENCODE, and one of the cell lines most commonly used for large-scale CRISPR/Cas9 gene-editing screens. Although the functional genomic and epigenomic characteristics of K562 are extensively studied, its genome sequence has never been comprehensively analyzed and higher-order structural features of its genome beyond its karyotype were only cursorily known. The high degree of aneuploidy in K562 renders traditional genome variant analysis methods challenging and partially ineffective. Correct and complete interpretation of the extensive functional genomics data from K562 requires an understanding of the cell line's genome sequence and genome structure. We performed very-deep short-insert whole-genome sequencing, mate-pair sequencing, linked-read sequencing, karyotyping and array CGH, and used a combination of novel and established computational methods to identify and catalog a wide spectrum of genome sequence variants and genome structural features in K562: copy numbers (CN) of chromosome segments, SNVs and Indels (allele frequencycorrected based on copy-number), phased haplotype blocks (N50 = 2.72 Mb), structural variants (SVs) including complex genomic rearrangements, and novel mobile element insertions. A large fraction of SVs were phased, sequence assembled, and experimentally validated. Many chromosomes show striking loss of heterozygosity. To demonstrate the utility of this knowledge, we re-analyzed K562 RNA-Seq and whole-genome bisulfite sequencing data to detect and phase allele-specific expression and DNA methylation patterns, respectively. We show examples where deeper insights into genomic regulatory complexity could be gained by taking knowledge of genomic structural contexts into account. Furthermore, we used the haplotype information to produce a phased CRISPR targeting map, i.e. a catalog of loci where CRISPR guide RNAs will bind in an allele-specific manner. This comprehensive whole-genome analysis serves as a resource for future studies that utilize K562 and as the basis of advanced analyses of the rich amounts of the functional genomics data produced by ENCODE for K562. It is also an example for advanced, integrated whole-genome sequence and structure analysis, beyond standard short-read/short-insert whole-genome sequencing, of human genomes in general and in particular of cancer genomes with large numbers of complex sequence alterations.

A deep learning approach to predict the impact of non-coding sequence variants on 3D chromatin structure

“…Additionally, the models were also tested with the data from chromosomes 7 and 8 of the cell line K562 5 . To increase the sensitivity of the models with mutations from patients, DNA sequences of boundaries were mutated with mutations from corresponding cell lines GM12878 22 and K562 23 for training, validating and testing. All datasets have approximately the same numbers of positive and negative samples and we used area under the receiver operating characteristic curve (AUROC) to measure the performance of the models.…”

Minimized double guide RNA libraries enable scale-limited CRISPR/Cas9 screens

“…In iPSCs, we observed an average 1.6-fold increase in estimated cell number for gRNAs targeting X, and a 3.2-fold increase in cell number for non-targeting gRNAs ( Figure 3C) compared to non-essential gene targeting guides. The effect in K562 cells was substantially smaller (1.5-fold increase in cells with non-targeting gRNAs, Figure 3D) likely reflecting the complex karyotype (Zhou et al 2019) and DNA repair activity of the K562 cell line. We further performed assays to compare survival and metabolic activity of cells five days post infection for the single gRNA Yusa 1.0 library, and the minimized double gRNA library.…”

Contribution of structural variation to genome structure: TAD fusion discovery and ranking

“…Furthermore, we also compared our distance based model for predicting contact frequencies between genomic pairs crossing several validated somatic deletions in K562 sample [50] and compared against the experimentally Hi-C matrix of the same sample [31]. The results of K562 comparison can be seen in Supplementary Figure S2 and Supplementary Table S2.…”