Version 1.0.5573

Released on October 12, 2022

Changes:

Switched to new procedure for computing peaks from HMM:
1. Estimate HMM posterior error probabilities
2. Pick candidate bins with relaxed posterior error probability, e.g. sqrt(fdr).
  This mitigates the problem of wide marks peaks split on strong fdrs.
3. Using gap merge bins into candidate islands.
4. Assign p-value to each island using based on combined p-values for blocks of consequent enriched bins.
  Each block is assigned P as average posterior log error probability for bins in blocks.
  50% top significant blocks scores are aggregated using length-weighted average as P for island.
5. Compute qvalues on islands p-values, filter by alpha.
Improved multistart initial NB states guess - use different params for each model
Safeguard against infinite failures as initial guess for NB, which effectively converts NB to Poisson
Reduced time required for peak calling
Show progress while computing peaks from model
Updated defaults values for SPAN peaks: fdr 0.05, gap 3
Updated defaults for SPAN model: total iterations 20, log-likelihood threshold 1
Added advanced SPAN model viewer
Bugfixes

Provide feedback