Dispersion and HMM
If Dispersion and HMM is the selected method, then the CNV tool first calculates the coverage ratios for each region. The tool then calculates the amount of dispersion (noise) for each region. The noise can be calculated automatically or manually. Finally, a Hidden Markov Model (HMM) uses the coverage ratio value and the amount of noise in each region to calculate a CNV classification (Duplication, Normal, Deletion, or Uncalled) for each region. Two options are available for calculating the coverage ratios:
• Normalized counts – Selected by default. Ratios are based on read counts for each region with both samples normalized by a size factor.
• RPKM – Ratios are based on RPKM measurements, where the measurements are read counts that are normalized by region length and the total number of reads.