 |
 |
GEO help: Mouse over screen elements for information. |
|
| Status |
Public on Nov 05, 2018 |
| Title |
AK231_H3K36me3_ChIP-seq |
| Sample type |
SRA |
| |
|
| Source name |
adult glioblastoma tumour
|
| Organism |
Homo sapiens |
| Characteristics |
subtype: IDH
gender: male
age: 56
survival_status: dead
os (overall survival, month): 8
progression: 1
pfs (progression-free survival, month): 4
chr7_gain: 0
chr10_loss: 1
chr19_gain: 0
chr20_gain: 0
idh1 mutation status: R132H
idh2 mutation status: wt
egfr_amplification: 0
pten_deletion: 0
mdm2_amplification: 0
mdm4_amplification: 0
pdgfra_amplification: 0
cdkn2a_b_deletion: 1
cdk4_amplification: 0
met_amplification: 0
molecule subtype: genomic DNA and associated chromatin
chip antibody: H3K36me3
chip antibody vendor: Active Motif
chip antibody cat.#: AM#61101
chip antibody lot #: 7
|
| Extracted molecule |
genomic DNA |
| Extraction protocol |
H3K27Ac (AM#39133, Active Motif), H3K4me1 (AM#38297, Active Motif), H3K4me3 (AM#39159, Active Motif), H3K9me3 (AM#39161, Active Motif), H3K27me3 (#07-449, Millipore) and H3K36me3 (AM#61101, Active Motif) ChIP library preparation of GB samples was performed at Active Motif according to proprietary methods.
Libraries were multiplexed so that all libraries for each individual IP were sequenced on 1-4 lanes using the Illumina HiSeq 2000 platform.
|
| |
|
| Library strategy |
ChIP-Seq |
| Library source |
genomic |
| Library selection |
ChIP |
| Instrument model |
Illumina HiSeq 2000 |
| |
|
| Description |
processed data file:
AK231_ChromHMM_18states.bed
AK231_ChromHMM_simplified_7states.bed
IDH_ChromHMM_18states.bed
IDH_ChromHMM_simplified_7states.bed
|
| Data processing |
ChIP-seq datasets were processed using a custom pipeline implemented in Snakemake. Briefly, reads were trimmed using the Trimgalore tool (https://github.com/FelixKrueger/TrimGalore) and aligned using Bowtie with standard parameters. Duplicates and multi-mapping reads were removed using the samtools package and the XS flag in the bam files. Input control (WGS) and corresponding IP datasets were scaled using the SES method and converted into a bigwig track using the bamCompare tool of the deepTools2 suite. Peaks were called using the callpeak mode in MACS2 (https://github.com/taoliu/MACS) for broad and narrow peaks. In addition, SICER was used to call peaks using the gap 600 and window 200 parameters. Various QC parameters (FRiP, PCR bottleneck coefficient, cross-strand correlation) were determined according to the ENCODE guidelines. In addition, visual QC was performed using the signal profile at TSS of annotated genes and the fingerprint method from the DeepTools2 suite.
Chromatin segmentation was defined using the ChromHMM tool. ChIP-seq data (H3K27ac, H3K27me3, H3K36me3, H3K4me1, H3K4me3, H3K9me3) and WGS data (input control) were binarized using ChromHMM’s “BinarizeBam” command. The Epigenome Roadmap 18-state model was used to segment the genome of each sample. The consensus state for a subtype is defined as the state with the highest frequency in a given segment, and a minimum frequency of 50%.
The simplified 7-state model was defined by mapping the 18 states as follows: 1-4 to TSS (Tss), 5-6 to Transcription (Tx), 7-11 and 15 to Enhancer (Enh), 12-13 to Heterochromatin/Repeats (Het/Rpts), 14 to Bivalent TSS (TSSBiv), 16-17 to Polycomb Repressed (ReprPC), 18 to Quiescent (Qui).
Subtype superenhancer analysis: the union of H3K27ac peaks for each subtype’s samples were used as input regions for the ROSE2 superenhancer analysis pipeline, stitching together regions within 12.5kbp. Sample H3K27ac signal was calculated using ‘bigWigAverageOverBed’ (v2), and enhancers were ranked by the subtype average enrichment. SEs were defined using the default parameters for ROSE2.
GBM superenhancer landscape: the pan-GBM SE landscape was defined as the reduced union of the 4 subtype SE lists. Enrichment of H3K27ac for each sample was calculated with ‘bigWigAverageOverBed’. z-scores of the log10-transformed H3K27ac signal was used to hierarchically cluster SEs using 1-Pearson correlation as the distance and the ‘Ward.D2’ method. Target genes were assigned as follows: for each SE, we computed the Spearman correlation of the H3K27ac signal in the SE regions across all samples with the expression of all genes within a region of +/- 500kb, and selected the gene showing the highest correlation as being the most likely target gene. We additionally filtered using TADs by requiring the SE and the predicted target gene to be within a common TAD.
Subtype enhancer analysis: subtype enhancers were defined as regions where a H3K4me1 peak was called in at least 2 samples in that subtype falling more than 2kbp from a H3K4me3 peak, or a promoter otherwise. Regions closer than 10bp were merged, and those longer than 500bp were retained. Active enhancers were defined as regions where a H3K27ac peak was also called, otherwise the region was defined as a poised enhancer. Intensities of H3K27ac ChIPseq signal was calculated using ‘bigWigAverageOverBed’.
Identification of the GBM active enhancer landscape: The GBM active enhancer landscape was defined by taking the union of the 4 subtype active enhancer sets. The activity of enhancers was visualised in a heatmap using ‘ComplexHeatmap’ of the z-scores of the log2-transformed H3K27ac signal. Active enhancers were clustered on the average log2 H3K27ac signal in each subtype, using 1-Pearson correlation as the distance and the ‘Ward.D2’ hierarchical clustering method.
Subtype-activated enhancer analysis: subtype-activated active enhancers were identified using an ANOVA test on the average subtype log2-transformed H3K27ac signal. Enhancers with a BH-adjusted ANOVA p-value < 0.1 and a log2FC of H3K27ac signal > 1 were defined as activated in a subtype.
Assignment of target genes to enhancers: The ‘InTAD’ R package (unpub., v0.1.1) was used to identify target genes of active enhancers. Briefly, the correlation of log2-transformed H3K27ac signal and protein-coding gene (Gencode v19 annotation) expression (log2 RPKM+1) was calculated within ESC-defined topologically associated domains for all possible enhancer-gene interactions. Genes were allowed to overlap multiple TADs (option “selMaxTadOvlp=TRUE” in the combineInTAD function). The target for each enhancer was defined as follows: the closest gene (distance to TSS) with correlation > 0.4; otherwise, the gene with the highest correlation.
genome build: hg19
processed data files format and content: ChromHMM results for each sample are contained in two bedfiles: the full 18-state and the reduced 7-state model. For each IPed factor, 3 files are included: a bigWig coverage file and two peak call bedfiles. Additionally, subtype consensus files are included for the ChromHMM models (both 18 and 7 states). Superenhancer analysis: the individual subtype SE lists (SUBTYPE_subtype_H3K27ac_superenhancers.bed) as well as the full landscape (GB_H3K27ac_superenhancers.bed), with subtype-specific SEs. Active enhancer analysis: individual subtype active enhancer lists (SUBTYPE_enhancers_annotated_with_histone_signal) are indluced, as well as the full landscape (GB_active_enhancers.bed) as well as subtype-specific lists. The full enhancer matrix, consisting of this full landscape with annotated targeted genes, is also included (GB_active_enhancers_matrix_with_target_gene_predictions.csv).
|
| |
|
| Submission date |
Oct 24, 2018 |
| Last update date |
Nov 15, 2018 |
| Contact name |
Bernhard Radlwimmer |
| E-mail(s) |
b.radlwimmer@dkfz-heidelberg.de
|
| Organization name |
Deutsches Krebsforschungszentrum / German National Cancer Research Centre
|
| Department |
Department of Molecular Genetics
|
| Street address |
Im Neuenheimer Feld 280
|
| City |
Heidelberg |
| ZIP/Postal code |
69120 |
| Country |
Germany |
| |
|
| Platform ID |
GPL11154 |
| Series (2) |
| GSE121719 |
Glioblastoma epigenome profiling identifies SOX10 as a master regulator of molecular tumour subtype - tumour ChIPseq data |
| GSE121723 |
Glioblastoma epigenome profiling identifies SOX10 as a master regulator of molecular tumour subtype |
|
| Relations |
| BioSample |
SAMN10285553 |
| Supplementary file |
Size |
Download |
File type/resource |
| GSM3444548_AK231_H3K36me3_W200-G600-FDR0.01-island.bed.gz |
306.3 Kb |
(ftp)(http) |
BED |
| GSM3444548_AK231_H3K36me3_coverage_SES_subtract.bigWig |
132.7 Mb |
(ftp)(http) |
BIGWIG |
| GSM3444548_AK231_H3K36me3_peaks.narrowPeak.bed.gz |
1.1 Mb |
(ftp)(http) |
BED |
| Raw data not provided for this record |
| Processed data provided as supplementary file |
| Processed data are available on Series record |
|
|
|
|
 |
