Table 1.
Genome-wide real data evaluation based on false-positive rate (FPR), false discovery rate (FDR) and recover rate (RR)
| Dataset | Method | FPR (%) | FDR (%) | RR (%) | |||||
|---|---|---|---|---|---|---|---|---|---|
| IMR90 | HMRF-Bayesian | 0.52 | 15.60 | 42.30 | |||||
| IMR90 | AFC | 0.60 | 18.40 | 41.10 | |||||
| IMR90 | Fit-Hi-C | 0.64 | 19.20 | 41.20 | |||||
| IMR90 + TNF-α | HMRF-Bayesian | 0.83 | 18.50 | 55.40 | |||||
| IMR90 + TNF-α | AFC | 0.98 | 22.40 | 52.90 | |||||
| IMR90 + TNF-α | Fit-Hi-C | 1.00 | 22.50 | 53.30 | |||||
| Dataset | Method | FPR | FDR | RR | |||||
|
| |||||||||
| Split1 | HMRF | 0.84 | 19.10 | 55.90 | |||||
| Split1 | AFC | 0.97 | 22.60 | 54.00 | |||||
| Split1 | Fit-Hi-C | 1.03 | 23.50 | 53.70 | |||||
| Split2 | HMRF | 0.47 | 15.20 | 41.30 | |||||
| Split2 | AFC | 0.56 | 18.60 | 39.90 | |||||
| Split2 | Fit-Hi-C | 0.58 | 18.80 | 40.30 | |||||
Assuming calling result for the combined dataset is the true peak pattern, we summarized the following measures for 1432 domains, i.e. genome-wide. We reported the genome-wide average of false-positive rate (FPR), false discovery rate (FDR) and recovery rate (RR) by the HMRF-Bayesian method, AFC method and Fit-Hi-C for both IMR90 before TNF-α treatment and IMR90 after TNF-α treatment. We found that the HMRF-Bayesian method has better performance than AFC method and Fit-Hi-C