CNV-HiCHIP Average Plot and Peak Calling
Data
Data Considered
| type | file |
|---|---|
| Input | MG63.3_input_GSE74230.bam |
| ChIP | MG63-GFP_H3K27ac_020823_NovoG.bam |
| HiChIP | MG63_H3K27ac_OsteoAQuA.bam |
Average Plots
Selected regions by CNV >= 1
Average Plots at each selected region
Number of Regions
| counts | regions |
|---|---|
| 311 | input_gt1seg.bed |
| 311 | chip_gt1seg.bed |
| 280 | hichip_gt1seg.bed |
- ChIP tends to overestimate, except in HiChIP-only regions.
- HiChIP is similar to Input.
- HiChIP overestimates in HiChIP regions.
Average Plots at exclusive reginos
Number of Regions
| counts | regions |
|---|---|
| 234 | input_or_chip_only.bed |
| 59 | hichip_only.bed |
| 194 | common.bed |
- In Input or ChIP regions, HiChIP predicts well, while ChIP tends to overestimate CNV.
- In the HiChIP only regions, HiChIP tends to overestimate.
Our Peak Calling
- Homer is too broad.
- MACS is too narrow and fails to detect peaks.
- Copywriter peaks overhang across segments, resulting in missed segments.
-
Fundamental Model
- Our method iteratively re-calls after removing those identified in the previous step.
Mismatching CNV regions
Our Peak Caller
| 1-iter | 3-iter |
|---|---|